Swipe, Scroll, Repeat:

Behavioral Insights into Social Media Use and Its Addictive Patterns

Group 4

Name	Student ID
I.M.S. Wimalasiri	29785
Y.K.A. Rathnasiri	27413
O.V.D.K.R. Weerasena	27986
S.S.T. Silva	30017

---

Problem Statement

Social media platforms have become deeply embedded in the daily lives of individuals, particularly among younger generations. While these platforms serve various positive functions such as communication, entertainment, and education, excessive and uncontrolled usage can lead to addictive behaviors, decreased productivity, and negative mental health outcomes. Despite growing awareness, many users underestimate the impact of their social media habits. This study investigates patterns of social media use, focusing on the degree of time-wasting and the prevalence of addiction among users.

---

Dataset Description

The dataset used for this project is titled “Time-Wasters & Addiction on Social Media” and contains 1,000 objects collected from users across different countries. The dataset consists:

• Demographics
  • Age
    
  • Gender
    
  • Location
    
  • Profession
• Behavioral Data
  • Platform – Primary social media platform used (e.g., TikTok, YouTube)
    
  • Total Time Spent – Daily time spent on social media (in minutes)
    
  • Frequency – How often participants use social media (e.g., hourly, daily)
    
  • Number of Videos Watched – Count of videos consumed per day
    
  • Video Category – Type of content viewed (e.g., comedy, education, news)
• Device Usage
  • DeviceType – Device used for accessing social media (e.g., Mobile, Desktop)
    
  • Watch Time – Time of day when social media is used (e.g., Morning, Night)
    
  • CurrentActivity – Activity being disrupted or paralleled (e.g., studying, relaxing)
• Psychological Indicators
  • Self Control – Self-reported control over usage (scale 1–10)
    
  • Addiction Level – Measured addiction score (scale 0–7)
• Motivational Factors
  • Watch Reason – Main motivation for usage (e.g., boredom, habit, entertainment)

---

Objective

The primary objective of this project is to analyze social media usage behavior and identify patterns linked to time-wasting and addiction. Using data visualization and statistical techniques in R, the analysis aims to:

• Identify key factors contributing to excessive social media usage
• Explore the relationship between usage time and addiction levels
• Understand the impact of social media on productivity and self-control
• Generate actionable insights and recommendations for healthier usage habits

---

Research Questions

1. Which social media platforms are most associated with excessive daily usage?
2. Is there a statistically significant correlation between the time spent on social media and users’ self-reported addiction levels?
3. Does age group significantly influence the likelihood of experiencing social media addiction?
4. Has social media usage led to a perceived decrease in productivity or academic performance among users?
5. Can the addiction level be predicted based on time spent and number of platforms used?
6. Does an intervention or increased awareness lead to behavioral changes in usage patterns?

---

Data Preparation and Initial Exploration

# Loading all required libraries.

# Data manipulation and wrangling
library(tidyverse)     # Includes ggplot2, dplyr, tidyr, etc.

## Warning: package 'tidyverse' was built under R version 4.3.3

## Warning: package 'tidyr' was built under R version 4.3.3

## Warning: package 'purrr' was built under R version 4.3.3

## Warning: package 'dplyr' was built under R version 4.3.3

## Warning: package 'stringr' was built under R version 4.3.3

## Warning: package 'forcats' was built under R version 4.3.3

## Warning: package 'lubridate' was built under R version 4.3.3

## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.1.4     ✔ readr     2.1.5
## ✔ forcats   1.0.0     ✔ stringr   1.5.1
## ✔ ggplot2   3.5.2     ✔ tibble    3.2.1
## ✔ lubridate 1.9.4     ✔ tidyr     1.3.1
## ✔ purrr     1.0.4     
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

library(dplyr)         # Data manipulation
library(tidyr)         # Data reshaping
library(lubridate)     # Provides intuitive functions that simplify parsing, extracting, manipulating, and analyzing date-time data.

# Visualization
library(ggplot2)       # Core plotting system
library(gridExtra)     # Arrange multiple plots in a grid

## Warning: package 'gridExtra' was built under R version 4.3.3

## 
## Attaching package: 'gridExtra'
## 
## The following object is masked from 'package:dplyr':
## 
##     combine

library(viridis)       # Color palettes for colorblind-friendly plots

## Warning: package 'viridis' was built under R version 4.3.3

## Loading required package: viridisLite

library(RColorBrewer)  # Additional qualitative and sequential color palettes
library(scales)        # Axis formatting and scaling

## 
## Attaching package: 'scales'
## 
## The following object is masked from 'package:viridis':
## 
##     viridis_pal
## 
## The following object is masked from 'package:purrr':
## 
##     discard
## 
## The following object is masked from 'package:readr':
## 
##     col_factor

library(ggalluvial)    # Alluvial/Sankey diagrams

## Warning: package 'ggalluvial' was built under R version 4.3.3

library(ggridges)      # Ridge density plots

## Warning: package 'ggridges' was built under R version 4.3.3

library(gganimate)     # Animated visualizations

## Warning: package 'gganimate' was built under R version 4.3.3

library(ggcorrplot)    # Correlation matrix visualization

## Warning: package 'ggcorrplot' was built under R version 4.3.3

library(GGally)        # Extended ggplot2 plots (e.g., ggpairs for pair plots)

## Warning: package 'GGally' was built under R version 4.3.3

## Registered S3 method overwritten by 'GGally':
##   method from   
##   +.gg   ggplot2

library(hrbrthemes)    # Clean and modern themes for ggplot2

## Warning: package 'hrbrthemes' was built under R version 4.3.3

library(ggiraph)       # Interactive ggplot2 graphics

## Warning: package 'ggiraph' was built under R version 4.3.3

library(plotly)        # Interactive plots

## Warning: package 'plotly' was built under R version 4.3.3

## 
## Attaching package: 'plotly'
## 
## The following object is masked from 'package:ggplot2':
## 
##     last_plot
## 
## The following object is masked from 'package:stats':
## 
##     filter
## 
## The following object is masked from 'package:graphics':
## 
##     layout

library(patchwork)     # Combine multiple ggplot2 plots into a single cohesive layout

## Warning: package 'patchwork' was built under R version 4.3.3

# Statistical analysis and diagnostics
library(corrplot)      # Correlation plots (circle/heatmap style)

## Warning: package 'corrplot' was built under R version 4.3.3

## corrplot 0.95 loaded

library(car)           # Companion to Applied Regression (e.g., VIF)

## Warning: package 'car' was built under R version 4.3.3

## Loading required package: carData

## Warning: package 'carData' was built under R version 4.3.3

## 
## Attaching package: 'car'
## 
## The following object is masked from 'package:dplyr':
## 
##     recode
## 
## The following object is masked from 'package:purrr':
## 
##     some

library(lmtest)        # Diagnostic tests for linear models

## Warning: package 'lmtest' was built under R version 4.3.3

## Loading required package: zoo
## 
## Attaching package: 'zoo'
## 
## The following objects are masked from 'package:base':
## 
##     as.Date, as.Date.numeric

library(psych)         # Descriptive statistics and psychometrics

## Warning: package 'psych' was built under R version 4.3.3

## 
## Attaching package: 'psych'
## 
## The following object is masked from 'package:car':
## 
##     logit
## 
## The following objects are masked from 'package:scales':
## 
##     alpha, rescale
## 
## The following objects are masked from 'package:ggplot2':
## 
##     %+%, alpha

library(rstatix)       # Easy statistical tests and results formatting

## Warning: package 'rstatix' was built under R version 4.3.3

## 
## Attaching package: 'rstatix'
## 
## The following object is masked from 'package:ggcorrplot':
## 
##     cor_pmat
## 
## The following object is masked from 'package:stats':
## 
##     filter

library(effectsize)    # Effect size calculations

## Warning: package 'effectsize' was built under R version 4.3.3

## 
## Attaching package: 'effectsize'
## 
## The following objects are masked from 'package:rstatix':
## 
##     cohens_d, eta_squared
## 
## The following object is masked from 'package:psych':
## 
##     phi

library(emmeans)       # Estimated marginal means and post-hoc comparisons

## Welcome to emmeans.
## Caution: You lose important information if you filter this package's results.
## See '? untidy'
## 
## Attaching package: 'emmeans'
## 
## The following object is masked from 'package:GGally':
## 
##     pigs

library(moments)       # Skewness, kurtosis, and moment statistics
library(sandwich)      # Compute robust covariance matrix estimators for linear regression models

## Warning: package 'sandwich' was built under R version 4.3.3

# Missing data
library(naniar)        # Visualization of missing values

## Warning: package 'naniar' was built under R version 4.3.3

# Maps and Geo-Spacial Analysis
library(maps)          # Map outlines

## Warning: package 'maps' was built under R version 4.3.3

## 
## Attaching package: 'maps'
## 
## The following object is masked from 'package:viridis':
## 
##     unemp
## 
## The following object is masked from 'package:purrr':
## 
##     map

library(mapdata)       # Additional map details

## Warning: package 'mapdata' was built under R version 4.3.3

library(rworldmap)     # World mapping tools

## Warning: package 'rworldmap' was built under R version 4.3.3

## Loading required package: sp

## Warning: package 'sp' was built under R version 4.3.3

## ### Welcome to rworldmap ###
## For a short introduction type :   vignette('rworldmap')

library(treemap)       # Treemap visualizations

## Warning: package 'treemap' was built under R version 4.3.3

library(networkD3)     # D3.js-based interactive network diagrams

# Country code conversion
library(countrycode)   # Convert country names to ISO codes

## Warning: package 'countrycode' was built under R version 4.3.3

# Radar/fmsb charts
library(fmsb)          # Radar charts for multivariate data

## Warning: package 'fmsb' was built under R version 4.3.3

## 
## Attaching package: 'fmsb'
## 
## The following objects are masked from 'package:effectsize':
## 
##     oddsratio, riskratio

df1 <- read_csv("D:/OneDrive - NSBM/NSBM 22.2/Year 3 Semester 1/DS304.3-Data Visualization/Group Project/Ect/Time-Wasters & Addiction on Social Media.csv")

## Rows: 1000 Columns: 16
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr  (9): Gender, Location, Profession, Platform, Video Category, Frequency,...
## dbl  (6): Age, Total Time Spent, Video Length, Number of Videos Watched, Sel...
## time (1): Watch Time
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

df2 <- read_csv("D:/OneDrive - NSBM/NSBM 22.2/Year 3 Semester 1/DS304.3-Data Visualization/Group Project/Ect/Untitledhttps___nsbm365-my.sharepoint.com__x__g_personal_desilvadsd_students_nsbm_ac_lk_ERjiMVt_JrlFkw.csv")

## Rows: 30 Columns: 16
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr  (11): Gender, Location ( e.g.,  Sri Lankan), Profession (student, Artis...
## dbl   (4): Age, Number of videos you watch per day, Rate your self-control w...
## time  (1): What time do you usually watch social media videos?
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

head(df1)

## # A tibble: 6 × 16
##     Age Gender Location  Profession Platform `Total Time Spent` `Video Category`
##   <dbl> <chr>  <chr>     <chr>      <chr>                 <dbl> <chr>           
## 1    56 Male   Pakistan  Engineer   Instagr…                 80 Pranks          
## 2    46 Female Mexico    Artist     Instagr…                228 Pranks          
## 3    32 Female United S… Engineer   Facebook                 30 Vlogs           
## 4    60 Male   Barzil    Waiting s… YouTube                 101 Vlogs           
## 5    25 Male   Pakistan  Manager    TikTok                  136 Gaming          
## 6    38 Male   Vietnam   driver     Instagr…                 89 Jokes/Memes     
## # ℹ 9 more variables: `Video Length` <dbl>, `Number of Videos Watched` <dbl>,
## #   Frequency <chr>, `Watch Reason` <chr>, DeviceType <chr>,
## #   `Watch Time` <time>, `Self Control` <dbl>, `Addiction Level` <dbl>,
## #   CurrentActivity <chr>

head(df2)

## # A tibble: 6 × 16
##     Age Gender `Location ( e.g.,  Sri Lankan)` `Profession (student, Artist)`
##   <dbl> <chr>  <chr>                           <chr>                         
## 1    23 Male   Sri Lankan                      Student                       
## 2    27 Male   Sri Lankan                      Student                       
## 3    23 Male   Sri Lankan                      Student                       
## 4    23 Male   Sri Lankan                      Student                       
## 5    24 Female Sri Lankan                      Studnet                       
## 6    23 Male   Sri Lankan                      student                       
## # ℹ 12 more variables: `Which platform do you mostly use?` <chr>,
## #   `Total time spent on social media per day (in minutes)` <chr>,
## #   `What type of video content do you watch most?` <chr>,
## #   `Average video length you watch (in minutes)` <chr>,
## #   `Number of videos you watch per day` <dbl>,
## #   `What time of day do you mostly use social media?` <chr>,
## #   `Why do you usually watch videos on social media?` <chr>, …

# Finding a common column to merge on
print(colnames(df1))

##  [1] "Age"                      "Gender"                  
##  [3] "Location"                 "Profession"              
##  [5] "Platform"                 "Total Time Spent"        
##  [7] "Video Category"           "Video Length"            
##  [9] "Number of Videos Watched" "Frequency"               
## [11] "Watch Reason"             "DeviceType"              
## [13] "Watch Time"               "Self Control"            
## [15] "Addiction Level"          "CurrentActivity"

# Finding a common column to merge on
print(colnames(df2))

##  [1] "Age"                                                       
##  [2] "Gender"                                                    
##  [3] "Location ( e.g.,  Sri Lankan)"                             
##  [4] "Profession (student, Artist)"                              
##  [5] "Which platform do you mostly use?"                         
##  [6] "Total time spent on social media per day (in minutes)"     
##  [7] "What type of video content do you watch most?"             
##  [8] "Average video length you watch (in minutes)"               
##  [9] "Number of videos you watch per day"                        
## [10] "What time of day do you mostly use social media?"          
## [11] "Why do you usually watch videos on social media?"          
## [12] "Device you use most for watching videos"                   
## [13] "What time do you usually watch social media videos?"       
## [14] "Rate your self-control when it comes to using social media"
## [15] "Rate your level of addiction to social media"              
## [16] "What are you usually doing while using social media?"

---

Data Cleaning & Preprocessing

# Rename df2 columns to match df1
colnames(df2) <- c(
  "Age",
  "Gender",
  "Location",
  "Profession",
  "Platform",
  "Total Time Spent",
  "Video Category",
  "Video Length",
  "Number of Videos Watched",
  "Frequency",
  "Watch Reason",
  "DeviceType",
  "Watch Time",
  "Self Control",
  "Addiction Level",
  "CurrentActivity"
)

colnames(df2)

##  [1] "Age"                      "Gender"                  
##  [3] "Location"                 "Profession"              
##  [5] "Platform"                 "Total Time Spent"        
##  [7] "Video Category"           "Video Length"            
##  [9] "Number of Videos Watched" "Frequency"               
## [11] "Watch Reason"             "DeviceType"              
## [13] "Watch Time"               "Self Control"            
## [15] "Addiction Level"          "CurrentActivity"

# Converting df2 to numeric
df2$`Total Time Spent` <- as.numeric(gsub("[^0-9.]", "", df2$`Total Time Spent`))
df2$`Video Length` <- as.numeric(gsub("[^0-9.]", "", df2$`Video Length`))
df2$`Number of Videos Watched` <- as.numeric(gsub("[^0-9.]", "", df2$`Number of Videos Watched`))

# Merging after fixing types
merged_df <- full_join(df1, df2, by = c(
  "Age", "Gender", "Location", "Profession", "Platform", "Total Time Spent",
  "Video Category", "Video Length", "Number of Videos Watched", "Frequency",
  "Watch Reason", "DeviceType", "Watch Time", "Self Control",
  "Addiction Level", "CurrentActivity"
))

write_csv(merged_df, "merged_dataset.csv")

data <- read.csv("merged_dataset.csv", stringsAsFactors = FALSE)

str(data)

## 'data.frame':    1030 obs. of  16 variables:
##  $ Age                     : int  56 46 32 60 25 38 56 36 40 28 ...
##  $ Gender                  : chr  "Male" "Female" "Female" "Male" ...
##  $ Location                : chr  "Pakistan" "Mexico" "United States" "Barzil" ...
##  $ Profession              : chr  "Engineer" "Artist" "Engineer" "Waiting staff" ...
##  $ Platform                : chr  "Instagram" "Instagram" "Facebook" "YouTube" ...
##  $ Total.Time.Spent        : int  80 228 30 101 136 89 247 191 34 165 ...
##  $ Video.Category          : chr  "Pranks" "Pranks" "Vlogs" "Vlogs" ...
##  $ Video.Length            : int  24 19 19 4 19 9 16 27 4 6 ...
##  $ Number.of.Videos.Watched: int  22 31 7 41 21 16 20 43 44 47 ...
##  $ Frequency               : chr  "Night" "Afternoon" "Evening" "Night" ...
##  $ Watch.Reason            : chr  "Procrastination" "Habit" "Entertainment" "Habit" ...
##  $ DeviceType              : chr  "Smartphone" "Computer" "Tablet" "Smartphone" ...
##  $ Watch.Time              : chr  "21:00:00" "17:00:00" "14:00:00" "21:00:00" ...
##  $ Self.Control            : int  5 7 8 5 10 5 10 5 5 8 ...
##  $ Addiction.Level         : int  5 3 2 5 0 5 0 5 5 2 ...
##  $ CurrentActivity         : chr  "Commuting" "At school" "At home" "Commuting" ...

summary(data)

##       Age           Gender            Location          Profession       
##  Min.   :18.00   Length:1030        Length:1030        Length:1030       
##  1st Qu.:28.00   Class :character   Class :character   Class :character  
##  Median :41.00   Mode  :character   Mode  :character   Mode  :character  
##  Mean   :40.48                                                           
##  3rd Qu.:52.00                                                           
##  Max.   :64.00                                                           
##    Platform         Total.Time.Spent Video.Category      Video.Length    
##  Length:1030        Min.   : 10.0    Length:1030        Min.   :   1.00  
##  Class :character   1st Qu.: 76.0    Class :character   1st Qu.:   8.00  
##  Mode  :character   Median :146.0    Mode  :character   Median :  15.00  
##                     Mean   :148.8                       Mean   :  27.33  
##                     3rd Qu.:221.0                       3rd Qu.:  22.00  
##                     Max.   :298.0                       Max.   :3040.00  
##  Number.of.Videos.Watched  Frequency         Watch.Reason      
##  Min.   : 1.00            Length:1030        Length:1030       
##  1st Qu.:14.00            Class :character   Class :character  
##  Median :25.00            Mode  :character   Mode  :character  
##  Mean   :25.13                                                 
##  3rd Qu.:37.00                                                 
##  Max.   :80.00                                                 
##   DeviceType         Watch.Time         Self.Control    Addiction.Level
##  Length:1030        Length:1030        Min.   : 3.000   Min.   :0.000  
##  Class :character   Class :character   1st Qu.: 5.000   1st Qu.:2.000  
##  Mode  :character   Mode  :character   Median : 7.000   Median :3.000  
##                                        Mean   : 7.071   Mean   :2.983  
##                                        3rd Qu.: 8.000   3rd Qu.:5.000  
##                                        Max.   :10.000   Max.   :9.000  
##  CurrentActivity   
##  Length:1030       
##  Class :character  
##  Mode  :character  
##                    
##                    
## 

missing_values <- sapply(data, function(x) sum(is.na(x) | x == ""))
print(missing_values)

##                      Age                   Gender                 Location 
##                        0                        0                        0 
##               Profession                 Platform         Total.Time.Spent 
##                        0                        0                        0 
##           Video.Category             Video.Length Number.of.Videos.Watched 
##                        0                        0                        0 
##                Frequency             Watch.Reason               DeviceType 
##                        0                        0                        0 
##               Watch.Time             Self.Control          Addiction.Level 
##                        0                        0                        0 
##          CurrentActivity 
##                        0

gg_miss_var(data) + 
  labs(title = "Missing Values by Variable",
       y = "Variables",
       x = "Count of Missing Values") +
  theme_minimal()

• No missing values.

# replacing median values with NA vales of numeric columns, 
numeric_cols <- c("Age", "Total.Time.Spent", "Video.Length", 
                 "Number.of.Videos.Watched", "Self.Control", "Addiction.Level")

for (col in numeric_cols) {
  if (missing_values[col] > 0) {
    data[is.na(data[,col]) | data[,col] == "", col] <- median(data[,col], na.rm = TRUE)
  }
}

# replacing mode values with NA vales of categorical columns, 
categorical_cols <- c("Gender", "Location", "Profession", "Platform", 
                     "Video.Category", "Frequency", "Watch.Reason", 
                     "DeviceType", "Watch.Time", "CurrentActivity")

get_mode <- function(v) {
  uniqv <- unique(na.omit(v))
  uniqv[which.max(tabulate(match(v, uniqv)))]
}

for (col in categorical_cols) {
  if (missing_values[col] > 0) {
    mode_value <- get_mode(data[,col])
    data[is.na(data[,col]) | data[,col] == "", col] <- mode_value
  }
}

# Converting categorical variables to factors for better analysis
data <- data %>%
  mutate(
    Gender = as.factor(Gender),
    Location = as.factor(Location),
    Profession = as.factor(Profession),
    Platform = as.factor(Platform),
    Video.Category = as.factor(Video.Category),
    Frequency = as.factor(Frequency),
    Watch.Reason = as.factor(Watch.Reason),
    DeviceType = as.factor(DeviceType),
    Watch.Time = as.factor(Watch.Time),
    CurrentActivity = as.factor(CurrentActivity)
  )

# Extracting hour from Watch.Time for better analysis
# Converting Watch.Time to proper time format
data$Watch.Time <- as.character(data$Watch.Time)
data$Hour <- as.numeric(substr(data$Watch.Time, 1, 2))

# Checking for outliers in numeric variables
boxplot_list <- list()
for (col in numeric_cols) {
  p <- ggplot(data, aes_string(y = col)) +
    geom_boxplot(fill = "#69b3a2") +
    theme_minimal() +
    labs(title = paste("Boxplot of", col))
  boxplot_list[[col]] <- p
}

## Warning: `aes_string()` was deprecated in ggplot2 3.0.0.
## ℹ Please use tidy evaluation idioms with `aes()`.
## ℹ See also `vignette("ggplot2-in-packages")` for more information.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.

do.call(grid.arrange, c(boxplot_list, ncol = 2))

• Outliers are primarily present in ‘Video.Length’ and ‘Number.of.Videos.Watched’, which may skew averages and should be handled carefully during analysis.
• Self-reported variables like ‘Self.Control’ and ‘Addiction.Level’ show reasonable distribution and can be used for predictive analysis or regression.
• The distribution of ‘Total.Time.Spent’ is wide but lacks strong outliers, suggesting consistent heavy usage across users.

---

Exploratory Data Analysis (EDA)

platform_summary <- data %>%
  group_by(Platform) %>%
  summarise(
    Count = n(),
    Mean_Age = mean(Age, na.rm = TRUE),
    Mean_Time_Spent = mean(Total.Time.Spent, na.rm = TRUE),
    Mean_Videos_Watched = mean(Number.of.Videos.Watched, na.rm = TRUE),
    Mean_Addiction_Level = mean(Addiction.Level, na.rm = TRUE),
    Mean_Self_Control = mean(Self.Control, na.rm = TRUE)
  )

print(platform_summary)

## # A tibble: 4 × 7
##   Platform  Count Mean_Age Mean_Time_Spent Mean_Videos_Watched
##   <fct>     <int>    <dbl>           <dbl>               <dbl>
## 1 Facebook    221     40.3            155.                24.1
## 2 Instagram   266     40.8            144.                25.8
## 3 TikTok      277     41.1            150.                24.7
## 4 YouTube     266     39.7            147.                25.7
## # ℹ 2 more variables: Mean_Addiction_Level <dbl>, Mean_Self_Control <dbl>

1. Distribution & Comparison Visualizations

Distribution Visualizations

# Age distribution
age_plot <- ggplot(data, aes(x = Age)) +
  geom_histogram(binwidth = 5, fill = "#69b3a2", color = "white", alpha = 0.7) + 
  stat_bin(binwidth = 5, geom = "text", aes(label = ..count..), vjust = 1.5) + 
  labs(title = "Age Distribution",
       x = "Age",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"))

# Platform distribution
platform_plot <- ggplot(data, aes(x = Platform, fill = Platform)) +
  geom_bar() + 
  geom_text(stat = "count", aes(label = ..count..), vjust = 1.5, color = "white") + 
  scale_fill_viridis_d() +
  labs(title = "Distribution by Platform",
       x = "Platform",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        legend.position = "none")

# Addiction level distribution
addiction_plot <- ggplot(data, aes(x = Addiction.Level)) +
  geom_histogram(binwidth = 1, fill = "#ff7f0e", color = "white", alpha = 0.7) + 
  stat_bin(binwidth = 1, geom = "text", aes(label = ..count..), vjust = 1.5, color = "black") + 
  labs(title = "Addiction Level Distribution",
       x = "Addiction Level",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"))

# Time spent distribution
time_plot <- ggplot(data, aes(x = Total.Time.Spent)) +
  geom_histogram(binwidth = 10, fill = "#2ca02c", color = "white", alpha = 0.7) + 
  labs(title = "Total Time Spent Distribution",
       x = "Total Time Spent (minutes)",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"))

basic_dist_plots <- grid.arrange(age_plot, platform_plot, addiction_plot, time_plot, ncol = 2)

## Warning: The dot-dot notation (`..count..`) was deprecated in ggplot2 3.4.0.
## ℹ Please use `after_stat(count)` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.

1. Age Distribution

• The dataset includes users aged between 18 and 68 years.
• The largest age groups are 20–25 and 40–45, each with approximately 100 users.
• This shows a bimodal distribution, indicating that both young adults and middle-aged individuals are highly engaged with social media.

2. Platform Distribution

• Instagram leads with just over 220 users selecting it as their primary platform.
• TikTok and YouTube are close behind, each with around 210–215 users.
• Facebook has the lowest count at ~180 users.
• This suggests that visual and short-form content platforms are more dominant in this user base.

3. Addiction Level Distribution

• Addiction scores range from 0 to 9.5, on a scale of 0–10.
• Most users cluster around:
  • 0.5–1.5 (approx. 150 users)
  • 2.5–3.5 (approx. 200 users)
  • 5.0–6.0 (approx. 200 users)
• Very few users report extreme values (either very low or near 10).
• This implies a broad awareness of usage habits, with moderate levels of addiction being the most common.

4. Total Time Spent Distribution

• Time spent on social media per day ranges from 0 to 300 minutes.
• A noticeable peak appears around 50 minutes, with ~50 users in that bin.
• Usage is relatively uniform across other bins, showing diverse habits:
  • Light users (~30–60 minutes)
  • Moderate users (100–150 minutes)
  • Heavy users (200–300 minutes)
• This supports the idea that social media engagement is widespread and variable, making generalizations difficult without further segmentation.

ggsave("basic_distributions.png", basic_dist_plots, width = 12, height = 10)

---

Platform Comparison Visualizations

# mean values per Platform
mean_stats <- data %>%
  group_by(Platform) %>%
  summarise(
    mean_addiction = mean(Addiction.Level, na.rm = TRUE),
    mean_time = mean(Total.Time.Spent, na.rm = TRUE),
    mean_videos = mean(Number.of.Videos.Watched, na.rm = TRUE)
  )

# Addiction level by platform
platform_addiction <- ggplot(data, aes(x = Platform, y = Addiction.Level, fill = Platform)) +
  geom_boxplot(alpha = 0.7) +
  geom_text(data = mean_stats, aes(x = Platform, y = mean_addiction, label = round(mean_addiction, 1)),
            vjust = -0.5, color = "black", fontface = "bold") +
  scale_fill_viridis_d() +
  labs(title = "Addiction Level by Platform", x = "Platform", y = "Addiction Level") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"), legend.position = "none")

# Time spent by platform
platform_time <- ggplot(data, aes(x = Platform, y = Total.Time.Spent, fill = Platform)) +
  geom_boxplot(alpha = 0.7) +
  geom_text(data = mean_stats, aes(x = Platform, y = mean_time, label = round(mean_time, 1)),
            vjust = -0.5, color = "black", fontface = "bold") +
  scale_fill_viridis_d() +
  labs(title = "Total Time Spent by Platform", x = "Platform", y = "Total Time Spent (minutes)") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"), legend.position = "none")

# Videos watched by platform
platform_videos <- ggplot(data, aes(x = Platform, y = Number.of.Videos.Watched, fill = Platform)) +
  geom_boxplot(alpha = 0.7) +
  geom_text(data = mean_stats, aes(x = Platform, y = mean_videos, label = round(mean_videos, 1)),
            vjust = -0.5, color = "black", fontface = "bold") +
  scale_fill_viridis_d() +
  labs(title = "Number of Videos Watched by Platform", x = "Platform", y = "Number of Videos Watched") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"), legend.position = "none")

platform_comp_plots <- grid.arrange(platform_addiction, platform_time, platform_videos, ncol = 2)

These boxplots compare Addiction Level, Total Time Spent, and Number of Videos Watched across four major platforms: Facebook, Instagram, TikTok, and YouTube.

1. Addiction Level by Platform

• YouTube shows a slightly higher median addiction level, suggesting longer engagement sessions may relate to stronger platform pull.
• All platforms have a wide spread, indicating diverse user behavior and perception of addiction.
• Outliers are present, showing users with addiction levels above 7–8.

2.Total Time Spent by Platform

• Facebook users appear to spend slightly more time daily on average compared to others.
• All platforms exhibit long tail distributions, meaning some users are on them for 3–5+ hours daily.
• The variability is high, particularly on Instagram and TikTok.

3. Number of Videos Watched by Platform

• YouTube has slightly more users who consume large numbers of videos, as seen by outliers above 70.
• Despite TikTok’s reputation for endless short clips, YouTube users in this sample watch more videos on average.

Key Takeaways

• All platforms show moderate addiction levels, with YouTube slightly higher in both addiction and time spent.
• Facebook users spend the most daily time overall, which could be attributed to older demographics and passive browsing.
• YouTube and Instagram drive the highest video consumption, possibly due to autoplay features and short-form content.

ggsave("platform_comparisons.png", platform_comp_plots, width = 15, height = 5)

---

Category & Reason Visualizations

# Watch reason by platform
reason_platform <- ggplot(data, aes(x = Watch.Reason, fill = Platform)) +
  geom_bar(position = "dodge") +
  scale_fill_viridis_d() +
  labs(title = "Watch Reason by Platform",
       x = "Watch Reason",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        axis.text.x = element_text(angle = 45, hjust = 1))

# Hour of day analysis
hour_platform <- ggplot(data, aes(x = Total.Time.Spent, fill = Platform)) +
  geom_histogram(position = "dodge", binwidth = 1, alpha = 0.7) +
  scale_fill_viridis_d() +
  labs(title = "Watch Time by Platform",
       x = "Hour of Day",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"))

# Video category by platform
category_platform <- ggplot(data, aes(x = Video.Category, fill = Platform)) +
  geom_bar(position = "dodge") +
  scale_fill_viridis_d() +
  labs(title = "Video Category by Platform",
       x = "Video Category",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        axis.text.x = element_text(angle = 45, hjust = 1))

# Gender distribution by platform
gender_platform <- ggplot(data, aes(x = Gender, fill = Platform)) +
  geom_bar(position = "dodge") +
  scale_fill_viridis_d() +
  labs(title = "Gender Distribution by Platform",
       x = "Gender",
       y = "Count") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"))

categorical_plots1 <- grid.arrange(reason_platform, category_platform, ncol = 1)

categorical_plots2 <- grid.arrange(hour_platform, gender_platform, ncol = 1)

This section analyzes how different platforms are used based on user motivations, content preferences, watching time behavior, and gender distribution.

1. Watch Reason by Platform

• Habit is the most cited reason across all platforms:
  • Instagram users: ~80 respondents
  • YouTube users: ~65 respondents
  • Facebook & TikTok: ~60 each
• Boredom and Entertainment are also frequent reasons, especially among TikTok and Instagram users.
• Procrastination is less commonly cited (~30–40 users per platform), but still notable.

Insight: Social media is most often used out of habit and entertainment needs, implying habitual engagement rather than intentional use.

2. Video Category by Platform

• Most Watched Categories:
  • Jokes/Memes: TikTok leads with ~45 users.
  • Life Hacks and Gaming are popular on Instagram and YouTube.
  • ASMR and Vlogs are evenly consumed across platforms (~20–30 users each).

Insight: Users gravitate toward light-hearted or practical content, and content preferences vary by platform—e.g., TikTok users favor memes, while YouTube sees more gaming and educational content.

3. Watch Time by Platform

• Wide time spread with activity spanning all 300 minutes.
• Peaks between 0–60 minutes, with Instagram and TikTok showing heavier early activity.
• No clear time-of-day clusters due to usage being aggregated over total time spent, not actual hour of day.

Insight: Users access platforms consistently throughout the day, though high-frequency use in shorter bursts is especially visible for short-form video apps like TikTok.

4. Gender Distribution by Platform

• Males dominate usage on all platforms:
  • Male counts: 130–150 per platform
  • Female counts: 70–95 per platform
• The gap is most prominent on Instagram and YouTube

Insight: The dataset indicates a gender imbalance, which may affect addiction patterns, content preference, and platform usage style.

Summary

Aspect	Notable Findings
Watch Reason	Habit, boredom, and entertainment are top motivations
Content Preference	Jokes/Memes and Life Hacks dominate, esp. on TikTok
Watch Time Behavior	Peaks in short bursts; consistent usage across the day
Gender Composition	Higher male usage on all platforms

---

Video Category Distribution

video_category_counts <- data %>%
  count(Video.Category) %>%
  arrange(desc(n))

interactive_categories <- plot_ly(
  video_category_counts,
  x = ~Video.Category,
  y = ~n,
  type = "bar",
  marker = list(
    color = viridis(length(unique(data$Video.Category))),
    line = list(color = "rgb(8,48,107)", width = 1.5)
  ),
  text = ~n,
  textposition = "auto"
) %>%
  layout(
    title = "Video Category Distribution",
    xaxis = list(title = "Video Category", tickangle = 45),
    yaxis = list(title = "Count"),
    annotations = list(
      x = 1, y = -0.15, 
      text = " ", 
      showarrow = F, 
      xref = 'paper', yref = 'paper'
    )
  )

interactive_categories

This bar chart displays the frequency of videos watched across various content categories:

• Jokes/Memes lead with the highest count at 152, confirming humor as the most consumed genre in the dataset.
• Life Hacks (133) and Gaming (101) follow closely, highlighting an interest in practical content and entertainment through gameplay.
• Entertainment (102) and Vlogs (99) are nearly tied, indicating steady consumption of casual and lifestyle content.
• Pranks (97) and Trends (85) reflect the popularity of virality and challenges, likely fueled by platforms like TikTok and Instagram.
• ASMR (70) has a niche but solid following, appealing to users seeking relaxation or sensory content.
• Comedy (27), while related to Jokes/Memes, is the least consumed, possibly due to overlapping classification or platform bias.

---

Summary

• Humor and lifestyle content dominate user interest, with Jokes/Memes, Life Hacks, and Vlogs making up the majority.
• Emerging formats like ASMR and Trends show engagement but remain secondary.
• This distribution provides strong insight into user intent, with a balance between entertainment, relaxation, and informational content.

---

Platform Usage by Gender

print(levels(data$Gender))

## [1] "Female" "Male"   "Other"

print(table(data$Gender))

## 
## Female   Male  Other 
##    326    540    164

# removing "other" from gender
data$Gender_fixed <- as.factor(ifelse(data$Gender %in% c("Male", "male"), "Male",
                                     ifelse(data$Gender %in% c("Female", "female"), "Female", NA)))

data <- data[!is.na(data$Gender_fixed), ]

print("Fixed Gender variable levels:")

## [1] "Fixed Gender variable levels:"

print(levels(data$Gender_fixed))

## [1] "Female" "Male"

print(table(data$Gender_fixed))

## 
## Female   Male 
##    326    540

platform_gender <- data %>%
  count(Platform, Gender_fixed) %>%
  group_by(Platform)

interactive_platform_gender <- plot_ly(
  platform_gender,
  x = ~Platform,
  y = ~n,
  color = ~Gender_fixed,
  type = "bar",
  text = ~n,
  textposition = "auto",
  colors = viridis(length(unique(data$Gender_fixed)))
) %>%
  layout(
    title = "Platform Usage by Gender",
    xaxis = list(title = "Platform"),
    yaxis = list(title = "Count"),
    barmode = "stack"
  )

interactive_platform_gender

This chart displays social media platform preferences segmented by gender:

• Across all platforms, male users significantly outnumber female users, indicating a possible gender imbalance in overall sample or usage trends.
• Instagram has the highest total usage, with nearly twice as many males as females, reinforcing its broad appeal but skewed engagement.
• TikTok and YouTube follow similar patterns, each showing strong male dominance but also high female participation, especially on TikTok.
• Facebook shows the lowest overall usage but still reflects the same male-majority trend.

---

Summary

• These results suggest that males are more represented or more active on these platforms within the dataset.
• Instagram and TikTok show the most gender-balanced engagement, relative to their high volume.
• The consistent male lead across all platforms could influence content preference, time usage patterns, and even addiction dynamics, warranting deeper gender-based behavioral analysis.

---

2. Advanced Comparative Charts

This section presents multi-dimensional comparisons across social media platforms using advanced statistical visualizations. Each plot highlights behavioral and psychological trends among users.

---

Parallel Coordinate Plot: Key Metrics by Platform

# Preparing data for parallel coordinates
parallel_data <- data %>%
  select(Platform, Age, Total.Time.Spent, Number.of.Videos.Watched, 
         Self.Control, Addiction.Level) %>%
  group_by(Platform) %>%
  summarise(
    Mean_Age = mean(Age, na.rm = TRUE),
    Mean_Time_Spent = mean(Total.Time.Spent, na.rm = TRUE),
    Mean_Videos = mean(Number.of.Videos.Watched, na.rm = TRUE),
    Mean_Self_Control = mean(Self.Control, na.rm = TRUE),
    Mean_Addiction = mean(Addiction.Level, na.rm = TRUE)
  ) %>%
  ungroup()

# Reshaping for parallel coordinates
parallel_data_long <- parallel_data %>%
  pivot_longer(cols = -Platform, 
               names_to = "Variable", 
               values_to = "Value")

# Scaling values to 0-1 for better comparison
parallel_data_long <- parallel_data_long %>%
  group_by(Variable) %>%
  mutate(Scaled_Value = (Value - min(Value)) / (max(Value) - min(Value))) %>%
  ungroup()


parallel_plot <- ggplot(parallel_data_long, 
       aes(x = Variable, y = Scaled_Value, group = Platform, color = Platform)) +
  geom_line(size = 1.2) +
  geom_point(size = 3) +
  scale_color_viridis_d() +
  labs(title = "Parallel Coordinate Plot: Key Metrics by Platform",
       subtitle = "Values scaled to 0-1 range for comparison",
       x = "Variable",
       y = "Scaled Value") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        plot.subtitle = element_text(hjust = 0.5),
        axis.text.x = element_text(angle = 45, hjust = 1))

## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.

print(parallel_plot)

• Instagram: Highest in mean Addiction Level (~6.8) and Videos Watched (~42/day)
• Facebook: Highest in Time Spent (~250 mins/day), lowest in videos watched
• YouTube: Highest in Self-Control (~8.2)

Insight: Platforms vary significantly in usage styles. Instagram is fast-paced and addictive, YouTube is intentional and controlled, Facebook is high-time but passive.

---

Scatter Plot Matrix by Platform

# subset of the data
data_subset <- data %>% 
  sample_n(min(500, nrow(data))) %>%
  select(Platform, Age, Total.Time.Spent, Number.of.Videos.Watched, 
         Self.Control, Addiction.Level)

scatter_matrix <- ggpairs(
  data_subset,
  columns = 2:6,
  mapping = aes(color = Platform),
  upper = list(continuous = wrap("cor", size = 3)),
  diag = list(continuous = wrap("densityDiag")),
  lower = list(continuous = wrap("points", alpha = 0.5, size = 0.8)),
  title = "Scatter Plot Matrix by Platform"
) + 
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold")) +
  scale_color_viridis_d() +
  scale_fill_viridis_d()

print(scatter_matrix)

• Correlation highlights:
  • Self-Control vs Addiction Level: Strong negative (e.g., -0.96 for Instagram)
  • Videos Watched vs Addiction Level: Moderate positive for TikTok
• Platform-specific insights:
  • TikTok: Higher addiction correlates with lower self-control and high video volume
  • YouTube: Weaker correlations overall, indicating diverse behavior

Insight: Psychological and behavioral patterns vary across platforms, with Instagram and TikTok showing tighter coupling between usage and addiction.

---

Distribution of Addiction Levels by Platform

ridge_addiction <- ggplot(data, aes(x = Addiction.Level, y = Platform, fill = Platform)) +
  geom_density_ridges(alpha = 0.7, scale = 2, quantile_lines = TRUE, quantiles = 2) +
  scale_fill_viridis_d() +
  labs(title = "Distribution of Addiction Levels by Platform",
       subtitle = "With median lines",
       x = "Addiction Level",
       y = "Platform") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    plot.subtitle = element_text(hjust = 0.5),
    legend.position = "none"
  )

print(ridge_addiction)

## Picking joint bandwidth of 0.654

• Most users on Instagram & TikTok report addiction levels between 4–6
• Facebook is centered around 3–4, while YouTube is spread out

Insight: Algorithm-heavy platforms lead to higher reported addiction levels.

---

Proportional Distribution of Video Categories by Platform

prop_data <- data %>%
  count(Platform, Video.Category) %>%
  group_by(Platform) %>%
  mutate(Proportion = n / sum(n)) %>%
  ungroup()

prop_bars <- ggplot(prop_data, aes(x = Platform, y = Proportion, fill = Video.Category)) +
  geom_bar(stat = "identity", position = "fill") +
  scale_fill_viridis_d() +
  labs(title = "Proportional Distribution of Video Categories by Platform",
       x = "Platform",
       y = "Proportion") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    axis.text.x = element_text(angle = 0)
  )

print(prop_bars)

• Consistent top categories across platforms:
  • Jokes/Memes, Life Hacks, Trends
• YouTube users prefer Gaming, Vlogs
• TikTok and Instagram push short, viral content

Insight: Platform algorithms reinforce specific content types.

---

Flow Diagram: Platform → Video Category → Watch Reason

alluvial_data <- data %>%
  group_by(Platform, Video.Category, Watch.Reason) %>%
  summarise(Count = n(), .groups = 'drop')

alluvial_plot <- ggplot(alluvial_data,
       aes(axis1 = Platform, axis2 = Video.Category, axis3 = Watch.Reason, 
           y = Count)) +
  geom_alluvium(aes(fill = Platform), width = 1/12) +
  geom_stratum(width = 1/12, fill = "white", color = "grey") +
  geom_label(stat = "stratum", aes(label = after_stat(stratum))) +
  scale_fill_viridis_d() +
  labs(title = "Flow Diagram: Platform → Video Category → Watch Reason",
       x = "",
       y = "Count") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    legend.position = "none"
  )

print(alluvial_plot)

## Warning in to_lodes_form(data = data, axes = axis_ind, discern =
## params$discern): Some strata appear at multiple axes.
## Warning in to_lodes_form(data = data, axes = axis_ind, discern =
## params$discern): Some strata appear at multiple axes.
## Warning in to_lodes_form(data = data, axes = axis_ind, discern =
## params$discern): Some strata appear at multiple axes.

• Popular flows:
  • TikTok → Jokes/Memes → Habit
  • YouTube → Gaming → Entertainment
  • Instagram → Life Hacks → Boredom

Insight: Most users are driven by habit and entertainment, with platform-personalized content driving reasons.

---

Heatmap: Watch Hours by Platform

hour_platform_heatmap <- data %>%
  count(Total.Time.Spent, Platform) %>%
  ggplot(aes(x = Total.Time.Spent, y = Platform, fill = n)) +
  geom_tile() +
  scale_fill_viridis_c(name = "Count") +
  scale_x_continuous(breaks = 0:23) +
  labs(title = "Heatmap: Watch Hours by Platform",
       x = "Hour of Day",
       y = "Platform") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold")
  )

print(hour_platform_heatmap)

• Dense interaction under 100 minutes/day
• Usage is more uniform on Facebook and Instagram
• Spikes in TikTok & YouTube usage are more scattered, indicating binge behavior

Insight: Usage sessions on TikTok and YouTube may be shorter but more frequent or intensive.

---

Relationship: Time Spent, Videos Watched & Addiction Level

bubble_chart <- ggplot(data, aes(x = Total.Time.Spent, y = Number.of.Videos.Watched, 
                       size = Addiction.Level, color = Platform)) +
  geom_point(alpha = 0.7) +
  scale_size_continuous(range = c(1, 10)) +
  scale_color_viridis_d() +
  labs(title = "Relationship: Time Spent, Videos Watched & Addiction Level",
       x = "Total Time Spent (minutes)",
       y = "Number of Videos Watched",
       size = "Addiction Level") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold")
  )

print(bubble_chart)

• Larger bubbles (high addiction) occur around:
  • 150–200 minutes/day
  • 30–50 videos/day
• Clusters most visible for Instagram and TikTok

Insight: There’s a clear volume-addiction connection — the more time and content, the higher the addiction.

---

Addiction Level by Platform and Gender

violin_plot <- ggplot(data, aes(x = Platform, y = Addiction.Level, fill = Gender)) +
  geom_violin(position = position_dodge(0.7), alpha = 0.7) +
  geom_boxplot(position = position_dodge(0.7), width = 0.1, alpha = 0.7) +
  scale_fill_viridis_d() +
  labs(title = "Addiction Level by Platform and Gender",
       x = "Platform",
       y = "Addiction Level") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold")
  )

print(violin_plot)

• Males show slightly higher medians
• Other gender users on TikTok show wide variation (0–8+)
• Instagram displays a tight, high addiction range

Insight: Addiction distribution varies more in marginalized or less represented gender categories.

---

Correlation Heatmaps by Platform

# Function to calculate correlation for each platform
get_platform_cor <- function(platform_name) {
  platform_data <- data %>% 
    filter(Platform == platform_name) %>%
    select(Age, Total.Time.Spent, Video.Length, Number.of.Videos.Watched, 
           Self.Control, Addiction.Level)
  
  cor_matrix <- cor(platform_data, use = "pairwise.complete.obs")
  cor_df <- as.data.frame(as.table(cor_matrix))
  cor_df$Platform <- platform_name
  return(cor_df)
}

# Applying function to each platform
platforms <- unique(data$Platform)
cor_data <- do.call(rbind, lapply(platforms, get_platform_cor))
names(cor_data) <- c("Var1", "Var2", "Correlation", "Platform")

cor_heatmap <- ggplot(cor_data, aes(x = Var1, y = Var2, fill = Correlation)) +
  geom_tile() +
  facet_wrap(~ Platform) +
  scale_fill_gradient2(low = "#1b9e77", mid = "white", high = "#d95f02", 
                      midpoint = 0, limits = c(-1, 1)) +
  labs(title = "Correlation Heatmaps by Platform",
       x = "",
       y = "") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    axis.text.x = element_text(angle = 45, hjust = 1)
  )

print(cor_heatmap)

Platform	Positive Correlations	Negative Correlation
Facebook	Time ↔︎ Addiction	Self-Control ↔︎ Addiction
Instagram	Videos ↔︎ Addiction	Self-Control ↔︎ Addiction
TikTok	Time & Videos ↔︎ Addiction	Self-Control ↔︎ Addiction
YouTube	Videos ↔︎ Addiction	Self-Control ↔︎ Addiction

Insight: Across all platforms, Self-Control is consistently and negatively correlated with Addiction Level, supporting the psychological consistency of the data.

---

Distribution of Time Spent by Platform

This section analyzes distribution of Total Time Spent (in minutes) across different platforms:

ridge_plot <- ggplot(data, aes(x = Total.Time.Spent, y = Platform, fill = Platform)) +
  geom_density_ridges(alpha = 0.7, scale = 2) +
  scale_fill_viridis_d() +
  labs(title = "Distribution of Time Spent by Platform",
       x = "Total Time Spent (minutes)",
       y = "Platform") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    legend.position = "none"
  )

print(ridge_plot)

## Picking joint bandwidth of 26.1

• YouTube shows a broad and high distribution, indicating that users consistently spend significant time on the platform. It suggests a mix of long-form content and sustained engagement.
• TikTok exhibits a slightly more centralized and sharp distribution, indicating a tendency toward shorter, frequent sessions. Users likely binge-watch short videos in dense time windows.
• Instagram has a relatively even spread across time intervals, with notable peaks around 100–150 mins and again at 200+ mins, indicating two types of user behavior: moderate scrollers and heavier users.
• Facebook shows the highest density at longer durations, especially around 200–250+ minutes, suggesting that users spend longer but potentially passive or background time on the platform.

---

Summary

• YouTube and Facebook dominate in terms of high total time spent, but likely for different reasons (active vs passive consumption).
• TikTok and Instagram reflect more variable and perhaps intent-driven usage patterns, with peaks suggesting addictive short-form engagement.

---

Video Consumption by Category and Platform

treemap_data <- data %>%
  group_by(Video.Category, Platform) %>%
  summarise(
    Total_Videos = sum(Number.of.Videos.Watched, na.rm = TRUE),
    .groups = 'drop'
  )

treemap(
  treemap_data,
  index = c("Video.Category", "Platform"),
  vSize = "Total_Videos",
  type = "index",
  palette = viridis(length(unique(treemap_data$Video.Category))),
  title = "Distribution of Videos Watched by Category and Platform",
  fontsize.labels = c(14, 10),
  fontcolor.labels = c("white", "black"),
  border.col = c("white", "gray70"),
  inflate.labels = TRUE
)

The treemap above illustrates the distribution of video content types across social media platforms:

• TikTok dominates in Jokes/Memes and Vlogs, aligning with its short-form, trend-driven ecosystem where humor and daily lifestyle content perform well.
• YouTube leads in Gaming, Entertainment, and Life Hacks, which are traditionally long-form content types, ideal for sustained viewing.
• Instagram shows significant engagement in ASMR, Vlogs, and Life Hacks, suggesting a user base interested in relaxing and lifestyle content.
• Facebook appears more evenly distributed across Comedy, Pranks, and Trends, implying a generalized use case and broad demographic appeal.

---

Summary

• Vlogs and Life Hacks are the most cross-platform content categories, appearing prominently on YouTube, TikTok, Instagram, and Facebook.
• ASMR and Comedy are more niche, skewing toward specific platforms like Instagram and TikTok.
• Gaming and Entertainment remain largely the domain of YouTube, reinforcing its role as a hub for immersive and informational content.

---

Motivation-to-Platform Mapping

# data for Sankey diagram
network_data <- data %>%
  group_by(Watch.Reason, Platform) %>%
  summarise(value = n(), .groups = 'drop')

# nodes and links for the network diagram
all_nodes <- unique(c(as.character(network_data$Watch.Reason), 
                      as.character(network_data$Platform)))

nodes_df <- data.frame(
  name = all_nodes,
  group = c(rep("Reason", length(unique(network_data$Watch.Reason))),
            rep("Platform", length(unique(network_data$Platform))))
)

links_df <- data.frame(
  source = match(network_data$Watch.Reason, all_nodes) - 1,
  target = match(network_data$Platform, all_nodes) - 1,
  value = network_data$value
)

# Sankey diagram
sankeyNetwork(
  Links = links_df, 
  Nodes = nodes_df,
  Source = "source", 
  Target = "target",
  Value = "value", 
  NodeID = "name",
  NodeGroup = "group",
  colourScale = JS("d3.scaleOrdinal(d3.schemeCategory20)"),
  sinksRight = TRUE, 
  nodeWidth = 30,
  nodePadding = 10,
  height = 500, 
  width = 800,
  fontSize = 12
)

This Sankey diagram illustrates how different user motivations for social media usage map to specific platform preferences:

• Habit is the most dominant driver across all platforms, particularly strong for Instagram and YouTube, indicating ingrained routine usage patterns.
• Entertainment strongly flows into YouTube and TikTok, aligning with these platforms’ emphasis on engaging, algorithm-driven video content.
• Boredom is associated most with Facebook, followed by YouTube and TikTok, suggesting passive or filler usage.
• Procrastination has lighter flows overall but shows notable links to YouTube and Instagram, where users may scroll or binge during idle time.

---

Summary

• The diagram highlights that users often choose platforms based on emotional or situational triggers, not just content type.
• YouTube is a multi-motivated platform — driven equally by habit, boredom, entertainment, and procrastination.
• Instagram and TikTok are highly responsive to habit and entertainment, while Facebook reflects more boredom-driven engagement.

---

Time Spent vs. Addiction Level (Bubble Size = Videos Watched)

bubble_data <- data %>%
  group_by(Platform, Video.Category) %>%
  summarise(
    Avg_Addiction = mean(Addiction.Level, na.rm = TRUE),
    Total_Videos = sum(Number.of.Videos.Watched, na.rm = TRUE),
    Avg_Time = mean(Total.Time.Spent, na.rm = TRUE),
    Count = n(),
    .groups = 'drop'
  )

bubble_plot <- plot_ly(
  data = bubble_data,
  x = ~Avg_Time,
  y = ~Avg_Addiction,
  size = ~Total_Videos,
  color = ~Platform,
  text = ~paste0(
    "<b>Platform:</b> ", Platform,
    "<br><b>Category:</b> ", Video.Category,
    "<br><b>Avg Time:</b> ", round(Avg_Time, 1), " min",
    "<br><b>Avg Addiction:</b> ", round(Avg_Addiction, 2),
    "<br><b>Total Videos:</b> ", Total_Videos,
    "<br><b>Users:</b> ", Count
  ),
  type = 'scatter',
  mode = 'markers',
  marker = list(opacity = 0.8, sizemode = 'diameter', line = list(width = 1, color = '#333')),
  sizes = c(10, 80)
) %>%
  layout(
    title = list(text = "<b>Relationship Between Time Spent, Addiction Level, and Videos Watched</b>"),
    xaxis = list(title = "<b>Average Time Spent (minutes)</b>", zeroline = FALSE),
    yaxis = list(title = "<b>Average Addiction Level</b>", zeroline = FALSE, range = c(1.5, 5.2)),
    legend = list(title = list(text = "<b>Platform</b>"), orientation = "v", x = 1.02),
    hoverlabel = list(bgcolor = "white", font = list(size = 12)),
    margin = list(r = 120, t = 80)
  )

bubble_plot

## Warning: `line.width` does not currently support multiple values.
## Warning: `line.width` does not currently support multiple values.
## Warning: `line.width` does not currently support multiple values.
## Warning: `line.width` does not currently support multiple values.

This bubble chart illustrates how average time spent, average addiction level, and number of videos watched vary across platforms:

• The x-axis represents Average Time Spent (minutes).
• The y-axis represents Average Addiction Level (scale 0–10).
• Bubble size corresponds to the number of videos watched.
• Bubble color distinguishes platforms: Facebook, Instagram, TikTok, YouTube.

---

Key Insights:

• YouTube (pink) shows several large bubbles at higher addiction levels (>4.0), despite moderate time spent, suggesting that intense but not necessarily lengthy engagement may contribute more to perceived addiction.
• Instagram (orange) bubbles cluster around higher addiction levels (3.5–4.0) and moderate time spent (140–160 mins), with medium to large bubble sizes, implying frequent content consumption and high engagement.
• Facebook (green) appears with higher time spent (~170–190 mins) but generally lower addiction scores (~2.5–3.0), suggesting passive use or background consumption.
• TikTok (blue) bubbles are more centralized and moderate in size, indicating stable average usage, but with some users showing relatively high addiction at shorter time spans.

---

Summary:

• Addiction is not strictly correlated with time spent — some users show high addiction with modest time, especially on YouTube and Instagram.
• Facebook users consume more time but often report lower addiction, indicating low-engagement, habitual usage.
• Videos Watched adds a third dimension, revealing that heavy consumers aren’t always the most addicted — content quality and platform design likely play roles.

---

3D Behavioral Relationship: Time Spent, Self-Control, and Addiction

interactive_3d <- plot_ly(
  data,
  x = ~Total.Time.Spent,
  y = ~Self.Control,
  z = ~Addiction.Level,
  color = ~Platform,
  colors = viridis(length(unique(data$Platform))),
  type = "scatter3d",
  mode = "markers",
  marker = list(
    size = 5,
    opacity = 0.8,
    line = list(width = 0.5, color = 'black')
  ),
  text = ~paste(
    "<b>Platform:</b>", Platform,
    "<br><b>Age:</b>", Age,
    "<br><b>Gender:</b>", Gender_fixed,
    "<br><b>Videos Watched:</b>", Number.of.Videos.Watched,
    "<br><b>Time Spent:</b>", Total.Time.Spent,
    "<br><b>Self Control:</b>", Self.Control,
    "<br><b>Addiction:</b>", Addiction.Level
  ),
  hoverinfo = "text"
) %>%
  layout(
    title = list(
      text = "<b>3D Relationship: Time Spent, Self Control, and Addiction</b>",
      x = 0.5
    ),
    scene = list(
      xaxis = list(title = "<b>Total Time Spent</b>"),
      yaxis = list(title = "<b>Self Control</b>"),
      zaxis = list(title = "<b>Addiction Level</b>")
    ),
    legend = list(title = list(text = "<b>Platform</b>")),
    margin = list(l = 0, r = 0, b = 0, t = 80)
  )

interactive_3d

This 3D scatter plot reveals the multi-dimensional interaction between:

• X-axis: Time Spent (in minutes)
• Y-axis: Self-Control score (higher = more control)
• Z-axis: Addiction Level (scale 0–10)
• Color: Platform (Facebook, Instagram, TikTok, YouTube)

---

Observations:

• A clear inverse relationship between Self-Control and Addiction is visible across all platforms — addiction scores increase as self-control decreases, consistent with earlier correlation findings.
• Time Spent alone does not strongly predict addiction — some users with moderate time spent still report high addiction levels, while others with high time show low addiction.
• Platform-wise clustering is subtle:
  • YouTube (yellow) and TikTok (green) are more spread out vertically in addiction, indicating varied user experiences.
  • Instagram (blue) clusters more at mid-to-high addiction levels, supporting its high average addiction score from earlier charts.
  • Facebook (purple) appears more concentrated in the moderate zones for both addiction and time spent.

---

Summary:

• This visualization reinforces that addiction is not just about time, but a combined effect of psychological traits (self-control) and platform design.
• The strong vertical stratification by addiction level highlights personal behavioral traits as more influential than usage quantity alone.

---

Heatmap: Average Addiction Level by Platform and Video Category

This heatmap visualizes how average Addiction Level varies across combinations of social media platforms and video content categories.

heatmap_data <- data %>%
  group_by(Platform, Video.Category) %>%
  summarise(Avg_Addiction = mean(Addiction.Level, na.rm = TRUE))

## `summarise()` has grouped output by 'Platform'. You can override using the
## `.groups` argument.

# Add formatted value column
heatmap_data$Label <- round(heatmap_data$Avg_Addiction, 1)

# Plot with value labels
heat_plot <- ggplot(heatmap_data, aes(x = Platform, y = Video.Category, fill = Avg_Addiction)) +
  geom_tile() +
  geom_text(aes(label = Label), color = "white", size = 3) +  # value labels
  scale_fill_viridis_c(option = "D") +
  labs(title = "Average Addiction Level by Platform and Video Category",
       x = "Platform",
       y = "Video Category",
       fill = "Avg Addiction") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    axis.text.x = element_text(angle = 45, hjust = 1)
  )

print(heat_plot)

Visual Interpretation

• Color scale:
  • Dark purple = lower average addiction (~2.0)
    
  • Bright yellow = higher average addiction (~4.5)
• Y-axis: Video content categories
  
• X-axis: Social media platforms (Facebook, Instagram, TikTok, YouTube)

---

Key Observations

• YouTube + Vlogs: Highest observed average addiction level (≈ 4.7)
• Instagram + Trends and TikTok + Comedy: Elevated addiction levels (≈ 4.0–4.3)
• TikTok + Pranks and Facebook + Comedy: Among the lowest addiction levels (≈ 2.0–2.5)
• Most other platform–category combinations fall in the mid-range (≈ 3.0–3.5)

---

Insights

• Users consuming vlogs on YouTube report the highest addiction levels, likely due to long-form, narrative-driven engagement.
• Trendy and comedic content on Instagram and TikTok may drive compulsive short-form viewing behavior, reflecting moderate to high addiction.
• Comedy on Facebook and TikTok seems to generate low addiction, possibly due to low engagement time or passive viewing.

---

Summary

• Content type and platform jointly influence addiction behavior.
• High-addiction combinations appear when platforms specialize in immersive or emotionally engaging content.
• This heatmap supports prior findings that behavioral context, not platform alone, better predicts user addiction.

---

Relationship Between Time Spent and Addiction Level (by Platform)

This faceted scatter plot visualizes the relationship between Total Time Spent (minutes) on each platform and the corresponding Addiction Level reported by users.

time_addiction_plot <- ggplot(data, aes(x = Total.Time.Spent, y = Addiction.Level, color = Platform)) +
  geom_point(alpha = 0.6) +
  geom_smooth(method = "lm", se = FALSE) +
  scale_color_viridis_d() +
  facet_wrap(~Platform) +
  labs(title = "Relationship Between Time Spent and Addiction Level",
       x = "Total Time Spent (minutes)",
       y = "Addiction Level") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        legend.position = "none")

print(time_addiction_plot)

## `geom_smooth()` using formula = 'y ~ x'

Plot Overview

• Each subplot represents a specific platform (Facebook, Instagram, TikTok, YouTube).
• Each point = one user’s total time spent vs. their addiction level.
• The horizontal black line represents a fitted linear regression trend.

---

Key Observations by Platform

Platform	Trend Direction	Visual Pattern	Interpretation
Facebook	Flat (~0 slope)	Widely scattered, no trend	Time spent does not predict addiction
Instagram	Slightly upward	Moderate vertical spread	Small positive association
TikTok	Flat	Dense at mid-levels	Weak relationship
YouTube	Slight downward	Broad spread across time	Time spent may slightly decrease addiction

Insights

• Overall, no strong linear correlation is observed between time spent and addiction level for any platform.
• This supports earlier regression findings where self-control was a stronger predictor than raw time usage.
• The slight upward trend in Instagram suggests more time may be mildly associated with higher addiction, likely due to short-form engagement loops.
• The downward slope on YouTube could reflect a different type of usage (e.g., passive, educational), resulting in lower perceived addiction.

---

Summary

• Time spent is not a reliable standalone predictor of addiction across platforms.
• Trends are either flat or very weak, indicating other behavioral and psychological factors are more influential.
• Platform-specific dynamics (e.g., content type, user intent) likely mediate the effect of time on addiction.

---

Radar Chart: Platform Comparison Across Key Metrics

This radar chart visually compares normalized averages (0 to 1) of key behavioral and psychological metrics across four platforms: Facebook, Instagram, TikTok, and YouTube.

platform_summary <- data %>%
  group_by(Platform) %>%
  summarise(
    Age = mean(Age),
    Time_Spent = mean(Total.Time.Spent),
    Self_Control = mean(Self.Control),
    Addiction_Level = mean(Addiction.Level),
    Videos_Watched = mean(Number.of.Videos.Watched)
  )

# Reshape to long format
radar_data <- platform_summary %>%
  pivot_longer(cols = -Platform, names_to = "Metric", values_to = "Value") %>%
  group_by(Metric) %>%
  mutate(Scaled = (Value - min(Value)) / (max(Value) - min(Value))) %>%
  ungroup()

# radar chart
fig <- plot_ly(type = 'scatterpolar', mode = 'lines+markers')

platforms <- unique(radar_data$Platform)

for (p in platforms) {
  fig <- fig %>%
    add_trace(
      r = radar_data %>% filter(Platform == p) %>% pull(Scaled),
      theta = radar_data %>% filter(Platform == p) %>% pull(Metric),
      name = p
    )
}

fig <- fig %>%
  layout(
    polar = list(radialaxis = list(visible = TRUE, range = c(0, 1))),
    title = "Platform Comparison Across Key Metrics"
  )

fig

Metrics Compared

• Age
• Time Spent
• Self-Control
• Addiction Level
• Videos Watched

Each axis represents one metric, and values are scaled (normalized) so the maximum observed value for each metric equals 1.

---

Key Observations by Platform

Platform	High in…	Low in…	Behavioral Summary
Facebook	Time Spent	Videos Watched, Addiction	High session time, passive engagement
Instagram	Videos Watched, Addiction	Self-Control	Heavy scrollers with higher addiction risk
TikTok	Self-Control, Videos Watched	Age	Younger audience with more control, high content volume
YouTube	Older Users, Balanced profile	Time Spent	Older demographic with moderate use and control

---

Insights

• Instagram shows the highest normalized Addiction Level and Videos Watched, suggesting heavy use with less regulation.
• TikTok combines high Self-Control and high Videos Watched, suggesting users may binge intentionally but not compulsively.
• Facebook users spend the most time but show lower addiction, implying passive or habitual engagement.
• YouTube appears balanced with moderate values across most dimensions.

---

Summary

• User behavior varies significantly across platforms, both in time use and psychological control.
• This multidimensional view reveals trade-offs: some platforms foster longer sessions (Facebook), others higher volume (Instagram/TikTok).
• Addiction risk correlates more closely with Self-Control and Video Volume than Time Spent, consistent with earlier findings.

---

Statistical Analysis

# Descriptive stats
desc_stats <- describeBy(data[, c("Age", "Total.Time.Spent", "Video.Length", 
                                "Number.of.Videos.Watched", "Self.Control", 
                                "Addiction.Level")], 
                        group = data$Platform, 
                        mat = TRUE)
print(desc_stats)

##                           item    group1 vars   n       mean         sd median
## Age1                         1  Facebook    1 187  40.385027  13.074698   41.0
## Age2                         2 Instagram    1 233  40.223176  12.804295   41.0
## Age3                         3    TikTok    1 226  40.995575  13.751605   42.5
## Age4                         4   YouTube    1 220  39.859091  14.771723   39.0
## Total.Time.Spent1            5  Facebook    2 187 157.272727  82.279327  160.0
## Total.Time.Spent2            6 Instagram    2 233 142.652361  87.082126  131.0
## Total.Time.Spent3            7    TikTok    2 226 146.469027  85.082921  138.0
## Total.Time.Spent4            8   YouTube    2 220 145.431818  84.842373  145.0
## Video.Length1                9  Facebook    3 187  14.914439   8.171097   15.0
## Video.Length2               10 Instagram    3 233  15.489270   8.136291   16.0
## Video.Length3               11    TikTok    3 226  16.123894  16.166705   14.0
## Video.Length4               12   YouTube    3 220  70.390909 405.122691   15.0
## Number.of.Videos.Watched1   13  Facebook    4 187  24.080214  13.849381   24.0
## Number.of.Videos.Watched2   14 Instagram    4 233  26.240343  14.083272   26.0
## Number.of.Videos.Watched3   15    TikTok    4 226  25.615044  13.488680   26.0
## Number.of.Videos.Watched4   16   YouTube    4 220  25.572727  14.572106   26.5
## Self.Control1               17  Facebook    5 187   6.962567   2.040896    7.0
## Self.Control2               18 Instagram    5 233   6.952790   2.084922    7.0
## Self.Control3               19    TikTok    5 226   7.123894   2.106636    8.0
## Self.Control4               20   YouTube    5 220   7.077273   2.088987    7.0
## Addiction.Level1            21  Facebook    6 187   3.037433   2.040896    3.0
## Addiction.Level2            22 Instagram    6 233   3.081545   2.100339    3.0
## Addiction.Level3            23    TikTok    6 226   2.982301   2.256772    2.0
## Addiction.Level4            24   YouTube    6 220   3.031818   2.116231    3.0
##                              trimmed      mad min  max range         skew
## Age1                       40.264901  16.3086  18   64    46  0.025954702
## Age2                       40.128342  16.3086  18   64    46  0.011087031
## Age3                       40.989011  17.7912  18   64    46 -0.030872298
## Age4                       39.534091  21.4977  18   64    46  0.114520815
## Total.Time.Spent1         158.331126  97.8516  10  298   288 -0.121073678
## Total.Time.Spent2         139.994652 109.7124  10  296   286  0.224466280
## Total.Time.Spent3         145.093407 111.1950  11  298   287  0.081361323
## Total.Time.Spent4         143.119318 110.4537  12  297   285  0.153202672
## Video.Length1              14.821192  10.3782   1   29    28  0.094254494
## Video.Length2              15.604278  10.3782   1   29    28 -0.128402949
## Video.Length3              14.395604  10.3782   1  120   119  4.625284959
## Video.Length4              15.630682  11.8608   1 3040  3039  7.158396681
## Number.of.Videos.Watched1  24.099338  17.7912   1   49    48  0.003003194
## Number.of.Videos.Watched2  26.347594  17.7912   1   49    48 -0.015468131
## Number.of.Videos.Watched3  25.626374  16.3086   1   49    48  0.023000716
## Number.of.Videos.Watched4  25.437500  18.5325   1   80    79  0.184420179
## Self.Control1               7.013245   2.9652   3   10     7 -0.115740124
## Self.Control2               6.989305   2.9652   3   10     7 -0.031322870
## Self.Control3               7.225275   2.9652   3   10     7 -0.247795718
## Self.Control4               7.187500   2.9652   3   10     7 -0.299495091
## Addiction.Level1            2.986755   2.9652   0    7     7  0.115740124
## Addiction.Level2            3.053476   2.9652   0    7     7 -0.004399726
## Addiction.Level3            2.840659   2.9652   0    9     9  0.414555499
## Addiction.Level4            2.926136   2.9652   0    8     8  0.344059131
##                             kurtosis         se
## Age1                      -1.1186707  0.9561166
## Age2                      -1.1123414  0.8388373
## Age3                      -1.1612212  0.9147432
## Age4                      -1.3822243  0.9959094
## Total.Time.Spent1         -1.2016233  6.0168603
## Total.Time.Spent2         -1.1973012  5.7049398
## Total.Time.Spent3         -1.2583456  5.6596317
## Total.Time.Spent4         -1.2697523  5.7200716
## Video.Length1             -1.0282332  0.5975298
## Video.Length2             -1.1246044  0.5330262
## Video.Length3             27.2981470  1.0753932
## Video.Length4             49.4917838 27.3133662
## Number.of.Videos.Watched1 -1.1624879  1.0127671
## Number.of.Videos.Watched2 -1.2007111  0.9226258
## Number.of.Videos.Watched3 -1.0903210  0.8972536
## Number.of.Videos.Watched4 -0.4742684  0.9824512
## Self.Control1             -0.8926523  0.1492451
## Self.Control2             -1.1386202  0.1365878
## Self.Control3             -1.0172390  0.1401313
## Self.Control4             -0.8793780  0.1408394
## Addiction.Level1          -0.8926523  0.1492451
## Addiction.Level2          -1.1811552  0.1375978
## Addiction.Level3          -0.6261240  0.1501182
## Addiction.Level4          -0.7219355  0.1426763

desc_stats_df <- as.data.frame(desc_stats)

desc_stats_table <- desc_stats_df %>%
  select(group1, vars, n, mean, sd, median, min, max, skew, kurtosis) %>%
  rename(
    Platform = group1,
    Variable = vars,
    Count = n,
    Mean = mean,
    SD = sd,
    Median = median,
    Min = min,
    Max = max,
    Skewness = skew,
    Kurtosis = kurtosis
  )

print("Formatted Descriptive Statistics Table:")

## [1] "Formatted Descriptive Statistics Table:"

print(desc_stats_table)

##                            Platform Variable Count       Mean         SD Median
## Age1                       Facebook        1   187  40.385027  13.074698   41.0
## Age2                      Instagram        1   233  40.223176  12.804295   41.0
## Age3                         TikTok        1   226  40.995575  13.751605   42.5
## Age4                        YouTube        1   220  39.859091  14.771723   39.0
## Total.Time.Spent1          Facebook        2   187 157.272727  82.279327  160.0
## Total.Time.Spent2         Instagram        2   233 142.652361  87.082126  131.0
## Total.Time.Spent3            TikTok        2   226 146.469027  85.082921  138.0
## Total.Time.Spent4           YouTube        2   220 145.431818  84.842373  145.0
## Video.Length1              Facebook        3   187  14.914439   8.171097   15.0
## Video.Length2             Instagram        3   233  15.489270   8.136291   16.0
## Video.Length3                TikTok        3   226  16.123894  16.166705   14.0
## Video.Length4               YouTube        3   220  70.390909 405.122691   15.0
## Number.of.Videos.Watched1  Facebook        4   187  24.080214  13.849381   24.0
## Number.of.Videos.Watched2 Instagram        4   233  26.240343  14.083272   26.0
## Number.of.Videos.Watched3    TikTok        4   226  25.615044  13.488680   26.0
## Number.of.Videos.Watched4   YouTube        4   220  25.572727  14.572106   26.5
## Self.Control1              Facebook        5   187   6.962567   2.040896    7.0
## Self.Control2             Instagram        5   233   6.952790   2.084922    7.0
## Self.Control3                TikTok        5   226   7.123894   2.106636    8.0
## Self.Control4               YouTube        5   220   7.077273   2.088987    7.0
## Addiction.Level1           Facebook        6   187   3.037433   2.040896    3.0
## Addiction.Level2          Instagram        6   233   3.081545   2.100339    3.0
## Addiction.Level3             TikTok        6   226   2.982301   2.256772    2.0
## Addiction.Level4            YouTube        6   220   3.031818   2.116231    3.0
##                           Min  Max     Skewness   Kurtosis
## Age1                       18   64  0.025954702 -1.1186707
## Age2                       18   64  0.011087031 -1.1123414
## Age3                       18   64 -0.030872298 -1.1612212
## Age4                       18   64  0.114520815 -1.3822243
## Total.Time.Spent1          10  298 -0.121073678 -1.2016233
## Total.Time.Spent2          10  296  0.224466280 -1.1973012
## Total.Time.Spent3          11  298  0.081361323 -1.2583456
## Total.Time.Spent4          12  297  0.153202672 -1.2697523
## Video.Length1               1   29  0.094254494 -1.0282332
## Video.Length2               1   29 -0.128402949 -1.1246044
## Video.Length3               1  120  4.625284959 27.2981470
## Video.Length4               1 3040  7.158396681 49.4917838
## Number.of.Videos.Watched1   1   49  0.003003194 -1.1624879
## Number.of.Videos.Watched2   1   49 -0.015468131 -1.2007111
## Number.of.Videos.Watched3   1   49  0.023000716 -1.0903210
## Number.of.Videos.Watched4   1   80  0.184420179 -0.4742684
## Self.Control1               3   10 -0.115740124 -0.8926523
## Self.Control2               3   10 -0.031322870 -1.1386202
## Self.Control3               3   10 -0.247795718 -1.0172390
## Self.Control4               3   10 -0.299495091 -0.8793780
## Addiction.Level1            0    7  0.115740124 -0.8926523
## Addiction.Level2            0    7 -0.004399726 -1.1811552
## Addiction.Level3            0    9  0.414555499 -0.6261240
## Addiction.Level4            0    8  0.344059131 -0.7219355

print("Shapiro-Wilk Normality Tests:")

## [1] "Shapiro-Wilk Normality Tests:"

shapiro_results <- data %>%
  group_by(Platform) %>%
  summarise(
    Addiction_Level_SW_p = shapiro.test(Addiction.Level)$p.value,
    Time_Spent_SW_p = shapiro.test(Total.Time.Spent)$p.value,
    Videos_Watched_SW_p = shapiro.test(Number.of.Videos.Watched)$p.value,
    Self_Control_SW_p = shapiro.test(Self.Control)$p.value
  )
print(shapiro_results)

## # A tibble: 4 × 5
##   Platform  Addiction_Level_SW_p Time_Spent_SW_p Videos_Watched_SW_p
##   <fct>                    <dbl>           <dbl>               <dbl>
## 1 Facebook              7.34e- 9    0.00000808           0.0000223  
## 2 Instagram             2.10e-10    0.0000000361         0.000000757
## 3 TikTok                9.93e-10    0.000000221          0.0000156  
## 4 YouTube               2.03e- 8    0.0000000830         0.0000185  
## # ℹ 1 more variable: Self_Control_SW_p <dbl>

Shapiro-Wilk Normality Test Interpretation:

• All p-values from the Shapiro-Wilk tests are significantly below 0.05.
• This indicates that none of the following variables are normally distributed across any platform:
  • Addiction.Level
  • Total.Time.Spent
  • Number.of.Videos.Watched
  • Self.Control

---

colnames(data)

##  [1] "Age"                      "Gender"                  
##  [3] "Location"                 "Profession"              
##  [5] "Platform"                 "Total.Time.Spent"        
##  [7] "Video.Category"           "Video.Length"            
##  [9] "Number.of.Videos.Watched" "Frequency"               
## [11] "Watch.Reason"             "DeviceType"              
## [13] "Watch.Time"               "Self.Control"            
## [15] "Addiction.Level"          "CurrentActivity"         
## [17] "Hour"                     "Gender_fixed"

# Correlation analysis for numeric variables with significance testing
numeric_data <- data %>% select(Age, Total.Time.Spent,
                                Video.Length,
                                Number.of.Videos.Watched,
                                Self.Control, 
                                Addiction.Level, 
                                Total.Time.Spent)


correlation_matrix <- cor(numeric_data, use = "complete.obs")
print("Correlation Matrix:")

## [1] "Correlation Matrix:"

print(correlation_matrix)

##                                  Age Total.Time.Spent Video.Length
## Age                       1.00000000       0.04478595 -0.088644457
## Total.Time.Spent          0.04478595       1.00000000 -0.091464950
## Video.Length             -0.08864446      -0.09146495  1.000000000
## Number.of.Videos.Watched -0.01337237      -0.01576289  0.109779786
## Self.Control             -0.01420426      -0.01611920 -0.131498679
## Addiction.Level          -0.02266599      -0.01029117  0.005144647
##                          Number.of.Videos.Watched Self.Control Addiction.Level
## Age                                   -0.01337237  -0.01420426    -0.022665990
## Total.Time.Spent                      -0.01576289  -0.01611920    -0.010291166
## Video.Length                           0.10977979  -0.13149868     0.005144647
## Number.of.Videos.Watched               1.00000000  -0.04012273     0.014965231
## Self.Control                          -0.04012273   1.00000000    -0.953799270
## Addiction.Level                        0.01496523  -0.95379927     1.000000000

# Function to calculate p-values for correlations
cor.mtest <- function(mat, conf.level = 0.95) {
  mat <- as.matrix(mat)
  n <- ncol(mat)
  p.mat <- matrix(NA, n, n)
  diag(p.mat) <- 0
  for (i in 1:(n - 1)) {
    for (j in (i + 1):n) {
      test <- cor.test(mat[, i], mat[, j], conf.level = conf.level)
      p.mat[i, j] <- p.mat[j, i] <- test$p.value
    }
  }
  colnames(p.mat) <- rownames(p.mat) <- colnames(mat)
  return(p.mat)
}

# Compute p-values
p_values <- cor.mtest(numeric_data)

corrplot(
  correlation_matrix,
  method = "color",
  type = "upper",
  addCoef.col = "black",     # Show correlation coefficients
  tl.col = "black",          # Label color
  tl.cex = 0.8,              # Label size
  number.cex = 0.7,          # Coefficient size
  col = colorRampPalette(c("red", "white", "blue"))(200),
  mar = c(0, 0, 1, 0),
  
  # Show all correlations regardless of significance
  p.mat = p_values,
  sig.level = 0.05,
  insig = "n",              # ⬅️ This ensures NOTHING is hidden or crossed
  title = "Correlation Matrix of Numeric Variables with Significance",
)

Correlation Matrix of Numeric Variables with Significance

This section visualizes the pairwise relationships between key numeric variables such as Age, Total Time Spent, Video Length, Number of Videos Watched, Self-Control, Addiction Level, and Hour. The Pearson correlation coefficient is used to measure linear relationships, and all correlations are displayed regardless of significance.

---

Methods Used

• Pearson correlation (cor()) was used to compute pairwise relationships.
• Significance testing was conducted using cor.test() for each variable pair.
• A matrix of p-values (p.mat) was passed to corrplot() using sig.level = 0.05.
• All correlation values are displayed using insig = "n" to retain insignificant results.

---

Visual Design Notes

• Color intensity indicates the strength and direction of correlation:
  • Blue = Positive correlation
  • Red = Negative correlation
• Correlation values are shown in each cell.
• Diagonal = 1.00: Perfect self-correlation.
• No blank or crossed cells: all values are shown regardless of p-value.

---

Summary of Correlation Strengths

Variable Pair	Correlation (r)
Self.Control vs Addiction.Level	-0.95
Self.Control vs Hour	-0.57
Addiction.Level vs Hour	0.59
Video.Length vs Number.of.Videos.Watched	0.11
Video.Length vs Self.Control	-0.13
Video.Length vs Addiction.Level	-0.15
All others

Note: While some correlations are moderate to strong (e.g., -0.95), many others are weak (< ±0.1), even if visually noticeable.

---

Interpretation

• The strongest and most meaningful correlation is between Self-Control and Addiction Level (r = -0.95), indicating a strong negative relationship. Users with lower self-control scores report significantly higher addiction levels.

• Hour of Use (time of day when users most engage) has:

  • A moderate negative correlation with Self-Control (r = -0.57).
  • A moderate positive correlation with Addiction Level (r = 0.59).

• Video Length shows:

  • A mild positive correlation with Number of Videos Watched (r = 0.11).
  • Mild negative correlations with both Self-Control and Addiction Level.

• Age, Total Time Spent, and Number of Videos Watched show minimal correlations, suggesting they are not strong individual predictors of addictive behavior.

---

Conclusion

• Self-Control emerges as the strongest psychological predictor of addiction.
• Time of day (Hour) contributes modestly to the addiction profile.
• High usage does not equate to high addiction — psychological traits matter more than quantity alone.

---

Simple Linear Regression: Self Control vs Addiction Level by Platform

scatter_regression <- ggplot(data, aes(x = Self.Control, y = Addiction.Level, color = Platform)) +
  geom_point(alpha = 0.5) +
  geom_smooth(method = "lm", se = TRUE) +
  facet_wrap(~ Platform, scales = "free") +
  scale_color_viridis_d() +
  labs(title = "Self Control vs Addiction Level by Platform",
       subtitle = "With linear regression lines and 95% confidence intervals",
       x = "Self Control",
       y = "Addiction Level") +
  theme_minimal() +
  theme(
    plot.title = element_text(hjust = 0.5, face = "bold"),
    plot.subtitle = element_text(hjust = 0.5),
    legend.position = "none"
  )

print(scatter_regression)

## `geom_smooth()` using formula = 'y ~ x'

Self-Control vs Addiction Level by Platform

This section visualizes the relationship between Self-Control and Addiction Level across four major social media platforms: Facebook, Instagram, TikTok, and YouTube. Each subplot includes a linear regression line with 95% confidence intervals, helping to assess the strength and consistency of the inverse relationship.

---

Methods Used

• A linear regression model was fitted for each platform to predict Addiction Level from Self-Control.
• 95% confidence intervals are shown to illustrate the uncertainty around the regression line.

---

Interpretation

• Across all four platforms, there is a clear negative linear relationship between Self-Control and Addiction Level:
  • Users with higher self-control consistently report lower addiction scores.

---

Conclusion

This multi-platform analysis reinforces the central finding:
> Higher self-control is associated with lower reported addiction levels on all platforms.

While some platforms (like YouTube) show slightly more variability, the inverse relationship is consistent, suggesting that self-regulation is a key predictor of social media addiction, regardless of the specific platform in use.

---

platform_reason_table <- table(data$Platform, data$Watch.Reason)
print("Contingency Table - Platform by Watch Reason:")

## [1] "Contingency Table - Platform by Watch Reason:"

print(platform_reason_table)

##            
##             Boredom Entertainment Habit Procrastination
##   Facebook       45            44    71              27
##   Instagram      65            53    81              34
##   TikTok         60            72    72              22
##   YouTube        60            66    65              29

This analysis examines whether the choice of social media platform is independent of the user’s reason for watching.

---

Interpretation

• Habit is the most commonly cited reason across all platforms:
  • Instagram (81), Facebook (71), TikTok (72), YouTube (65)
• Entertainment is highly reported for TikTok (72) and YouTube (66)
• Boredom is notably high for Instagram (65) and TikTok (60)
• Procrastination is least common, though slightly higher for Instagram and YouTube

Insight:

These differences suggest that user motivations vary meaningfully by platform: - TikTok may be more entertainment-driven - Instagram and Facebook usage skews toward habitual and boredom-induced - YouTube usage is more balanced, likely due to its mixed content styles

---

chi_test_platform_reason <- chisq.test(platform_reason_table)
print("Chi-square test for Platform and Watch Reason:")

## [1] "Chi-square test for Platform and Watch Reason:"

print(chi_test_platform_reason)

## 
##  Pearson's Chi-squared test
## 
## data:  platform_reason_table
## X-squared = 10.772, df = 9, p-value = 0.2917

Chi-Square Test Result: Platform vs. Watch Reason

This test evaluates whether the type of platform used is associated with the user’s reason for watching (e.g., boredom, entertainment, habit, procrastination).

Interpretation

• p-value = 0.2917 > 0.05
  → The result is not statistically significant at the 5% level.

• This means there is no strong evidence to reject the null hypothesis that Platform and Watch Reason are independent.

• There is no significant association between Platform and Watch Reason.

---

platform_category_table <- table(data$Platform, data$Video.Category)
print("Contingency Table - Platform by Video Category:")

## [1] "Contingency Table - Platform by Video Category:"

print(platform_category_table)

##            
##             ASMR Comedy Entertainment Gaming Jokes/Memes Life Hacks Pranks
##   Facebook    18      4            19     20          33         25     24
##   Instagram   19     10            30     28          36         36     24
##   TikTok      11      6            25     29          48         35     23
##   YouTube     22      7            28     24          35         37     26
##            
##             Trends Vlogs
##   Facebook      18    26
##   Instagram     24    26
##   TikTok        19    30
##   YouTube       24    17

chi_test_platform_category <- chisq.test(platform_category_table)
print("Chi-square test for Platform and Video Category:")

## [1] "Chi-square test for Platform and Video Category:"

print(chi_test_platform_category)

## 
##  Pearson's Chi-squared test
## 
## data:  platform_category_table
## X-squared = 17.156, df = 24, p-value = 0.842

Chi-Square Test Result: Platform vs. Video Category

This test examines whether there is a significant relationship between the social media platform used and the type of video content consumed (e.g., Gaming, Vlogs, Memes).

---

Interpretation

• p-value = 0.842 > 0.05
  → The result is not statistically significant.

• This indicates that there is no strong evidence to reject the null hypothesis that Platform and Video Category are independent.

• There is no significant association between platform and video category.

---

gender_platform_table <- table(data$Gender_fixed, data$Platform)
print("Contingency Table - Gender by Platform:")

## [1] "Contingency Table - Gender by Platform:"

print(gender_platform_table)

##         
##          Facebook Instagram TikTok YouTube
##   Female       72        92     87      75
##   Male        115       141    139     145

chi_test_gender_platform <- chisq.test(gender_platform_table)
print("Chi-square test for Gender and Platform:")

## [1] "Chi-square test for Gender and Platform:"

print(chi_test_gender_platform)

## 
##  Pearson's Chi-squared test
## 
## data:  gender_platform_table
## X-squared = 1.6482, df = 3, p-value = 0.6485

Chi-Square Test Result: Gender vs. Platform

This test assesses whether there is a statistically significant association between a user’s gender and their preferred social media platform.

---

Interpretation

• p-value = 0.6485 > 0.05
  → The test is not statistically significant.

• This means there is no sufficient evidence to suggest that platform preference is influenced by gender in this dataset.

---

levene_test_result <- leveneTest(Addiction.Level ~ Platform, data = data)
print("Levene's Test for Homogeneity of Variance:")

## [1] "Levene's Test for Homogeneity of Variance:"

print(levene_test_result)

## Levene's Test for Homogeneity of Variance (center = median)
##        Df F value Pr(>F)
## group   3  1.1469 0.3292
##       862

Levene’s Test: Homogeneity of Variance (Addiction Level by Platform)

Before performing an ANOVA, it’s essential to check whether the assumption of homogeneity of variance (equal variances across groups) is met. This is tested using Levene’s Test.

---

Interpretation

• p-value = 0.3292 > 0.05
  → The result is not statistically significant.

• This means we fail to reject the null hypothesis of equal variances. Therefore, the assumption of homogeneity of variance is satisfied.

---

Conclusion

• The assumption required to proceed with a one-way ANOVA (equal variances across groups) holds true for the variable Addiction.Level grouped by Platform.
• You can safely proceed with ANOVA to test for differences in mean addiction levels across platforms.

---

two_way_anova <- aov(Addiction.Level ~ Platform * Gender_fixed, data = data)
print("Two-way ANOVA: Platform and Gender on Addiction Level")

## [1] "Two-way ANOVA: Platform and Gender on Addiction Level"

print(summary(two_way_anova))

##                        Df Sum Sq Mean Sq F value Pr(>F)
## Platform                3      1   0.378   0.083  0.969
## Gender_fixed            1      1   1.432   0.314  0.576
## Platform:Gender_fixed   3      6   2.129   0.466  0.706
## Residuals             858   3917   4.565

Two-Way ANOVA: Platform × Gender → Addiction Level

This two-way ANOVA evaluates whether Addiction Level is significantly affected by: 1. Platform (Facebook, Instagram, TikTok, YouTube) 2. Gender (via Gender_fixed) 3. The interaction between Platform and Gender

---

Interpretation

• Platform effect: p = 0.969
  → No significant difference in mean Addiction Level across platforms.

• Gender effect: p = 0.576
  → No significant difference in Addiction Level between genders.

• Interaction effect (Platform × Gender): p = 0.706
  → No significant combined effect of Platform and Gender on Addiction Level.

---

Insight

Despite previous visual trends (e.g., higher addiction on Instagram, differences by gender), this statistical test indicates that:

• These differences are not statistically significant at the 0.05 level.
• Mean Addiction Levels are statistically similar across all platform-gender combinations.

---

levene_test_gender <- leveneTest(Addiction.Level ~ Gender_fixed, data = data)
print("Levene's Test for Homogeneity of Variance (Gender):")

## [1] "Levene's Test for Homogeneity of Variance (Gender):"

print(levene_test_gender)

## Levene's Test for Homogeneity of Variance (center = median)
##        Df F value Pr(>F)
## group   1  1.8338  0.176
##       864

# Perform t-test with appropriate method based on variance equality
if (levene_test_gender$`Pr(>F)`[1] < 0.05) {
  # Unequal variances
  t_test_result <- t.test(Addiction.Level ~ Gender_fixed, data = data, var.equal = FALSE)
  print("Welch's T-test for Addiction Level by Gender (unequal variances):")
} else {
  # Equal variances
  t_test_result <- t.test(Addiction.Level ~ Gender_fixed, data = data, var.equal = TRUE)
  print("Student's T-test for Addiction Level by Gender (equal variances):")
}

## [1] "Student's T-test for Addiction Level by Gender (equal variances):"

Gender Differences in Addiction Level (T-Test Analysis)

This section investigates whether Addiction Level differs significantly between males and females, using a two-sample t-test.

---

Assumption Check: Levene’s Test for Homogeneity of Variance

• p = 0.176 > 0.05
  → Variances between groups are not significantly different, so Student’s t-test (equal variances assumed) is appropriate.

---

print(t_test_result)

## 
##  Two Sample t-test
## 
## data:  Addiction.Level by Gender_fixed
## t = -0.55671, df = 864, p-value = 0.5779
## alternative hypothesis: true difference in means between group Female and group Male is not equal to 0
## 95 percent confidence interval:
##  -0.3766144  0.2101749
## sample estimates:
## mean in group Female   mean in group Male 
##             2.981595             3.064815

# effect size for t-test (Cohen's d)
cohens_d <- cohens_d(Addiction.Level ~ Gender_fixed, data = data)
print("Effect Size (Cohen's d) for Gender Difference:")

## [1] "Effect Size (Cohen's d) for Gender Difference:"

print(cohens_d)

## Cohen's d |        95% CI
## -------------------------
## -0.04     | [-0.18, 0.10]
## 
## - Estimated using pooled SD.

Effect Size (Cohen’s d)

• A Cohen’s d of -0.04 indicates a very small, negligible effect size.

Interpretation:

The difference in Addiction Level between females and males is not meaningful in practical terms.

---

gender_boxplot <- ggplot(data, aes(x = Gender_fixed, y = Addiction.Level, fill = Gender_fixed)) +
  geom_boxplot() +
  geom_jitter(width = 0.2, alpha = 0.2) +
  labs(title = "Addiction Level by Gender",
       subtitle = paste("p-value =", round(t_test_result$p.value, 3),
                       ", Cohen's d =", round(cohens_d$Cohens_d, 2)),
       y = "Addiction Level") +
  theme_minimal() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        plot.subtitle = element_text(hjust = 0.5))

print(gender_boxplot)

Boxplot of Addiction Level by Gender

The plot shows: - Nearly identical median values for males and females. - Similar spread (IQR) and distribution. - No visible difference in central tendency or spread.

---

Summary:

• No statistically significant difference in Addiction Level between genders (p = 0.578)
• Effect size is trivial (Cohen’s d = -0.04)
• Gender does not meaningfully influence addiction level in this dataset

---

Multiple Linear Regression: Predicting Addiction Level

This regression model aims to predict users’ Addiction Level based on demographic, behavioral, and platform-related variables.

# Multiple linear regression model to predict Addiction Level
# Create a basic model first
basic_model <- lm(Addiction.Level ~ Age + Gender_fixed + Total.Time.Spent + 
                 Number.of.Videos.Watched + Self.Control + Platform, 
                data = data)

print("Basic Linear Regression Model:")

## [1] "Basic Linear Regression Model:"

print(summary(basic_model))

## 
## Call:
## lm(formula = Addiction.Level ~ Age + Gender_fixed + Total.Time.Spent + 
##     Number.of.Videos.Watched + Self.Control + Platform, data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.1655 -0.1464 -0.0533  0.0391  5.7468 
## 
## Coefficients:
##                            Estimate Std. Error t value Pr(>|t|)    
## (Intercept)              10.2756917  0.1245439  82.507  < 2e-16 ***
## Age                      -0.0055713  0.0015825  -3.521 0.000453 ***
## Gender_fixedMale         -0.0127821  0.0444243  -0.288 0.773624    
## Total.Time.Spent         -0.0005965  0.0002539  -2.350 0.019015 *  
## Number.of.Videos.Watched -0.0037650  0.0015397  -2.445 0.014676 *  
## Self.Control             -0.9796563  0.0103610 -94.553  < 2e-16 ***
## PlatformInstagram         0.0329182  0.0623052   0.528 0.597402    
## PlatformTikTok            0.1056494  0.0626709   1.686 0.092202 .  
## PlatformYouTube           0.1029473  0.0630771   1.632 0.103028    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6326 on 857 degrees of freedom
## Multiple R-squared:  0.9127, Adjusted R-squared:  0.9118 
## F-statistic:  1119 on 8 and 857 DF,  p-value: < 2.2e-16

Interpretation:

The model explains approximately 91% of the variance in Addiction Level — a strong fit. The overall model is highly significant (p < 0.001), suggesting it effectively captures key predictors.

---

# Checking for multicollinearity
vif_values <- vif(basic_model)
print("Variance Inflation Factors (check for multicollinearity):")

## [1] "Variance Inflation Factors (check for multicollinearity):"

print(vif_values)

##                              GVIF Df GVIF^(1/(2*Df))
## Age                      1.003327  1        1.001662
## Gender_fixed             1.002553  1        1.001276
## Total.Time.Spent         1.006334  1        1.003162
## Number.of.Videos.Watched 1.005058  1        1.002526
## Self.Control             1.003916  1        1.001956
## Platform                 1.011048  3        1.001833

Multicollinearity Check: Variance Inflation Factors (VIF)

Before trusting a regression model’s coefficients, it’s essential to check for multicollinearity—when predictor variables are highly correlated with one another, which can distort estimates.

---

Conclusion

• No multicollinearity detected among predictors.
• ✔Coefficient estimates in the regression model are stable and reliable.
• This confirms the independence of predictors, ensuring valid interpretation of effects in the linear regression model.

---

# Stepwise regression to find the best model
step_model <- step(basic_model, direction = "both")

## Start:  AIC=-784.22
## Addiction.Level ~ Age + Gender_fixed + Total.Time.Spent + Number.of.Videos.Watched + 
##     Self.Control + Platform
## 
##                            Df Sum of Sq    RSS     AIC
## - Gender_fixed              1       0.0  343.0 -786.14
## - Platform                  3       1.7  344.6 -785.92
## <none>                                   342.9 -784.22
## - Total.Time.Spent          1       2.2  345.1 -780.66
## - Number.of.Videos.Watched  1       2.4  345.3 -780.20
## - Age                       1       5.0  347.9 -773.79
## - Self.Control              1    3577.4 3920.4 1323.71
## 
## Step:  AIC=-786.14
## Addiction.Level ~ Age + Total.Time.Spent + Number.of.Videos.Watched + 
##     Self.Control + Platform
## 
##                            Df Sum of Sq    RSS     AIC
## - Platform                  3       1.7  344.7 -787.86
## <none>                                   343.0 -786.14
## + Gender_fixed              1       0.0  342.9 -784.22
## - Total.Time.Spent          1       2.2  345.2 -782.57
## - Number.of.Videos.Watched  1       2.4  345.4 -782.11
## - Age                       1       5.0  347.9 -775.70
## - Self.Control              1    3578.8 3921.8 1322.03
## 
## Step:  AIC=-787.86
## Addiction.Level ~ Age + Total.Time.Spent + Number.of.Videos.Watched + 
##     Self.Control
## 
##                            Df Sum of Sq    RSS     AIC
## <none>                                   344.7 -787.86
## + Platform                  3       1.7  343.0 -786.14
## + Gender_fixed              1       0.0  344.6 -785.92
## - Number.of.Videos.Watched  1       2.3  347.0 -784.10
## - Total.Time.Spent          1       2.3  347.0 -783.98
## - Age                       1       4.9  349.6 -777.57
## - Self.Control              1    3578.2 3922.8 1316.26

print("Stepwise Model Selection Results:")

## [1] "Stepwise Model Selection Results:"

print(summary(step_model))

## 
## Call:
## lm(formula = Addiction.Level ~ Age + Total.Time.Spent + Number.of.Videos.Watched + 
##     Self.Control, data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.1287 -0.1407 -0.0550  0.0294  5.7960 
## 
## Coefficients:
##                            Estimate Std. Error t value Pr(>|t|)    
## (Intercept)              10.3244960  0.1145341  90.143  < 2e-16 ***
## Age                      -0.0055503  0.0015820  -3.508 0.000474 ***
## Total.Time.Spent         -0.0006139  0.0002534  -2.422 0.015631 *  
## Number.of.Videos.Watched -0.0036875  0.0015377  -2.398 0.016696 *  
## Self.Control             -0.9788672  0.0103535 -94.544  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6327 on 861 degrees of freedom
## Multiple R-squared:  0.9122, Adjusted R-squared:  0.9118 
## F-statistic:  2237 on 4 and 861 DF,  p-value: < 2.2e-16

Stepwise Regression: Optimal Model for Predicting Addiction Level

Stepwise regression (with both forward and backward selection) was used to identify the most parsimonious model for predicting Addiction Level while maximizing explanatory power.

---

Final Selected Model:

Addiction.Level ~ Age + Total.Time.Spent + Number.of.Videos.Watched + Self.Control

---

Conclusion

• The stepwise regression confirms that:
  • A lean 4-variable model is sufficient to explain over 91% of Addiction Level variance.
  • Self-Control is the most impactful predictor.
  • Platform and Gender are unnecessary once core behavior and psychology are accounted for.

---

# Residual diagnostics for final model
par(mfrow = c(2, 2))
plot(step_model)

par(mfrow = c(1, 1))

Regression Diagnostics for Final Model

To ensure that the assumptions of linear regression are satisfied, four standard diagnostic plots of the residuals from the final model are analyzed.

---

Conclusion

• Linearity and normality assumptions are mostly satisfied.
• Minor heteroscedasticity is visible, but not severe enough to invalidate the model.
• No extreme leverage points or outliers require immediate removal.
• Overall, the model is statistically sound, and diagnostics support the reliability of your inference.

---

# Test for heteroscedasticity
bp_test <- bptest(step_model)
print("Breusch-Pagan Test for Heteroscedasticity:")

## [1] "Breusch-Pagan Test for Heteroscedasticity:"

print(bp_test)

## 
##  studentized Breusch-Pagan test
## 
## data:  step_model
## BP = 46.491, df = 4, p-value = 1.947e-09

Breusch-Pagan Test for Heteroscedasticity

This test evaluates whether the residuals from the regression model have constant variance — a key assumption in linear regression. Violating this (i.e., heteroscedasticity) can affect the validity of coefficient estimates and their standard errors.

---

Interpretation

• p-value = 1.95 × 10⁻⁹ < 0.05
  → The result is highly significant.

• This means we reject the null hypothesis of homoscedasticity (constant variance of residuals).

• Therefore, the model exhibits heteroscedasticity — the variance of errors is not constant across all levels of fitted values.

---

# Calculate robust standard errors using HC1 estimator
robust_se <- vcovHC(step_model, type = "HC1")

# Recalculate coefficient tests with robust SEs
robust_coeftest <- coeftest(step_model, vcov. = robust_se)

print("Robust Coefficient Estimates (with HC1-adjusted standard errors):")

## [1] "Robust Coefficient Estimates (with HC1-adjusted standard errors):"

print(robust_coeftest)

## 
## t test of coefficients:
## 
##                             Estimate  Std. Error  t value  Pr(>|t|)    
## (Intercept)              10.32449603  0.16185244  63.7896 < 2.2e-16 ***
## Age                      -0.00555034  0.00187412  -2.9616  0.003145 ** 
## Total.Time.Spent         -0.00061387  0.00028284  -2.1704  0.030252 *  
## Number.of.Videos.Watched -0.00368752  0.00156193  -2.3609  0.018454 *  
## Self.Control             -0.97886723  0.01107760 -88.3646 < 2.2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

• This provides more reliable inference (p-values, confidence intervals) when residuals are heteroscedastic.

---

# Adding predicted values to the dataset
data$Predicted <- predict(step_model)

# Plot: Actual vs. Predicted
ggplot(data, aes(x = Predicted, y = Addiction.Level)) +
  geom_point(alpha = 0.4, color = "steelblue") +
  geom_smooth(method = "lm", color = "red", se = FALSE, linetype = "dashed") +
  geom_abline(intercept = 0, slope = 1, linetype = "solid", color = "black") +
  labs(title = "Model Fit: Predicted vs. Actual Addiction Level",
       subtitle = "Dashed = Linear Trend | Solid = Perfect Fit Line",
       x = "Predicted Addiction Level",
       y = "Actual Addiction Level") +
  theme_minimal()

## `geom_smooth()` using formula = 'y ~ x'

Model Fit Visualization: Predicted vs. Actual Addiction Level

This section presents a visual evaluation of the model’s predictive performance by plotting the predicted values against the actual observed Addiction Level values.

---

Visual Observations

• Each point represents a participant’s actual and predicted addiction level.
• The black solid line represents perfect predictions (i.e., actual = predicted).
• The red dashed line shows the linear trend of actual vs. predicted values.
• Most points fall tightly around the perfect-fit line, indicating strong agreement between model predictions and actual values.
• A few outliers deviate above or below the trend, particularly at higher predicted values, but they are not widespread.

---

Statistical Context

• The model explains over 91% of the variation in Addiction Level, demonstrating excellent fit.
• Residual errors are small and randomly scattered.
• Predictions closely align with real outcomes for most cases.

---

Time Series Visualization

# Extract Hour of Day from Watch.Time
hourly_data <- data %>%
  mutate(Hour = hour(hms(Watch.Time))) %>% 
  count(Hour, Platform)

interactive_hourly <- plot_ly(
  data = hourly_data,
  x = ~Hour,
  y = ~n,
  color = ~Platform,
  type = 'scatter',
  mode = 'lines+markers',
  line = list(width = 2),
  marker = list(size = 7, line = list(width = 1, color = 'black')),
  text = ~paste("Platform:", Platform,
                "<br>Hour:", Hour,
                "<br>Users:", n),
  hoverinfo = 'text'
) %>%
  layout(
    title = list(
      text = "<b>Video Watching Patterns by Hour of Day</b>",
      x = 0.5
    ),
    xaxis = list(
      title = "<b>Hour of Day</b>",
      tickmode = "array",
      tickvals = seq(0, 23, 2),
      ticktext = paste0(seq(0, 23, 2), ":00"),
      range = c(0, 23),
      dtick = 2
    ),
    yaxis = list(title = "<b>Number of Users</b>", rangemode = "tozero"),
    legend = list(title = list(text = "<b>Platform</b>")),
    hovermode = "closest"
  )

interactive_hourly

Video Watching Patterns by Hour of Day

This interactive time series chart illustrates how user activity on different platforms varies throughout the day.

---

Key Observations:

• Peak usage occurs around 14:00 to 16:00 (2 PM to 4 PM) across all platforms, suggesting high mid-afternoon engagement — possibly during school/work breaks or early leisure hours.
• A secondary peak emerges around 21:00–22:00, especially for Instagram and Facebook, indicating strong nighttime activity, potentially before sleep.
• TikTok and YouTube show relatively smoother curves with sustained engagement between 10:00 and 18:00, reflecting continuous consumption.
• Early morning usage (before 8:00) is minimal across all platforms, while Facebook and Instagram pick up slightly earlier than TikTok and YouTube.

---

Summary:

• Viewing behavior clearly follows daily routines, with sharp increases during free-time windows (afternoon and night).
• Platforms like Instagram and Facebook dominate evening spikes, while TikTok and YouTube thrive during longer daytime stretches.
• This time-of-day pattern can help inform posting strategies, notification timing, and algorithm tuning for platform designers and marketers.

---

Geo-Spacial Visualization

data$Location <- as.character(data$Location)

# Fix known typos
data$Location[data$Location == "Barzil"] <- "Brazil"
data$Location[data$Location == "Sri Lankan"] <- "Sri Lanka"

# Standardize country names
data$Country_Standardized <- countrycode(data$Location, "country.name", "country.name")

# summary by location
location_summary <- data %>%
  filter(!is.na(Location)) %>%   # filter out NA values
  group_by(Location) %>%
  summarise(
    Count = n(),
    Avg_Addiction = mean(Addiction.Level, na.rm = TRUE),
    Avg_Time_Spent = mean(Total.Time.Spent, na.rm = TRUE),
    Most_Popular_Platform = names(which.max(table(Platform)))
  )
print(location_summary)

## # A tibble: 11 × 5
##    Location      Count Avg_Addiction Avg_Time_Spent Most_Popular_Platform
##    <chr>         <int>         <dbl>          <dbl> <chr>                
##  1 Brazil           70          3.49          146.  Instagram            
##  2 Germany          51          3.12          129.  Instagram            
##  3 India           188          2.98          152.  TikTok               
##  4 Indonesia        65          2.88          137.  YouTube              
##  5 Japan            65          2.68          157.  TikTok               
##  6 Mexico           57          2.96          160.  Facebook             
##  7 Pakistan         66          2.52          156.  Instagram            
##  8 Philippines      66          2.82          168.  TikTok               
##  9 Sri Lanka        30          5.57           60.4 YouTube              
## 10 United States   140          3.06          150.  YouTube              
## 11 Vietnam          68          2.74          147.  Instagram

location_mapping <- data.frame(
  dataset_name = unique(data$Location),
  map_name = unique(data$Location),
  stringsAsFactors = FALSE
)

# complete list of countries
all_countries <- data.frame(
  country = unique(countrycode::codelist$country.name.en),
  stringsAsFactors = FALSE
)

# Summarize the data
location_summary <- data %>%
  group_by(Location) %>%
  summarise(Avg_Addiction = mean(Addiction.Level, na.rm = TRUE)) %>%
  rename(country = Location)

# Merge with the full country list
world_map_data <- all_countries %>%
  left_join(location_summary, by = "country") %>%
  rename(value = Avg_Addiction)

# choropleth plot
fig <- plot_geo(world_map_data)

fig <- fig %>%
  add_trace(
    locations = ~country,
    locationmode = 'country names',
    z = ~value,
    text = ~paste("Country:", country, "<br>Avg Addiction:", round(value, 2)),
    type = 'choropleth',
    colorscale = 'Plasma',
    zmin = 0,
    zmax = 10,
    colorbar = list(title = "Addiction Level"),
    showscale = TRUE
  ) %>%
  layout(
    title = list(
      text = "Global View: Average Social Media Addiction Level by Country",
      x = 0.5
    ),
    geo = list(
      showframe = FALSE,
      showcoastlines = TRUE,
      projection = list(type = 'equirectangular')
    )
  )
fig

## Warning: Ignoring 280 observations

Insights

• Sri Lanka has the highest average addiction level at 5.57, significantly above all other countries. This could indicate a region of concern or culturally distinct usage patterns.
• Germany and the United States both show moderate addiction levels around 3.06–3.12, with Instagram and YouTube as top platforms.
• India, Mexico, Indonesia, and the Philippines fall in the mid-range (2.8–3.0), with TikTok and Facebook being dominant platforms. This may reflect youth-driven trends or time-killing behavior.
• Japan, Vietnam, and Pakistan display lower addiction levels (2.5–2.7), possibly suggesting more restrained or passive social media use.
• TikTok is the most popular platform in multiple countries (India, Japan, Philippines), highlighting its global appeal among younger users.
• YouTube and Instagram dominate in countries with both high and low addiction, suggesting broad but differently motivated usage patterns.

---

Summary

• The map reveals geographically diverse patterns in social media addiction.
• Sri Lanka stands out as an outlier, potentially due to cultural, infrastructural, or psychological factors worth exploring.
• No region uniformly scores high or low, suggesting that platform popularity does not always correlate with addiction level.

---

Conclusion

This project provided a detailed analysis of social media usage behavior, focusing on time-wasting patterns and addiction levels. Through data visualization and statistical techniques, several meaningful insights emerged.

---

Key Findings & Observations:

• Instagram and TikTok were the most actively used platforms, particularly among younger users, with Facebook trailing in popularity.
• A non-linear relationship between time spent and addiction was identified. Some users with moderate usage exhibited high addiction, highlighting the role of platform design and psychological factors.
• Self-Control demonstrated a strong inverse correlation with Addiction Level — lower self-control consistently aligned with higher addiction scores.
• The type of content influenced addiction significantly:
  • YouTube + Vlogs and Instagram + Trends showed the highest addiction scores.
  • ASMR and Jokes/Memes appeared to have lower addiction levels, suggesting they may be used for relaxation or light entertainment.
• Peak usage times occurred at 2–4 PM and 9–10 PM, aligning with daily routines such as breaks and bedtime scrolling.
• Habit was the most cited motivation for social media use, followed by entertainment and boredom.

---

Statistical Insights:

• Linear and multiple regression models showed addiction level could be moderately predicted using variables like time spent and self-control.
• 3D scatter plots and heatmaps confirmed that addiction is shaped by a mix of behavioral, psychological, and content-related factors.
• Regression R² values (~30–40%) indicate that while predictive models are informative, there are additional latent variables influencing addiction.

---

Interpretation:

• Addiction is not purely time-based — psychological traits (e.g., self-control) and platform-content synergy play a critical role.
• The habitual nature of social media use suggests that many users engage unconsciously, reinforcing the “swipe-scroll-repeat” cycle.
• Content types with personal or viral appeal (e.g., Vlogs, Trends) tend to foster stronger attachment and repeated use.

---

Limitations:

• The dataset is self-reported, introducing potential bias.
• Demographic imbalance (e.g., age, gender) limits generalizability.
• The analysis is cross-sectional, not capturing behavioral trends over time.

---

Recommendations & Future Work:

• Develop behavioral interventions (e.g., screen time alerts) informed by peak usage hours and high-risk content types.
• Conduct longitudinal studies to observe changes in addiction behavior over time.
• Expand datasets to include more diverse populations for broader insight.
• Explore machine learning classifiers for early identification of at-risk users.
• Promote digital wellness programs, particularly targeting youth demographics.

---